Fiat Chrysler’s sales-reporting practices have put the company at the center of investigations amid allegations that it inflated car sales numbers for years. A new report from the company presents a supposedly improved counting method, which—after being applied to previous sales reports—reveals that Fiat Chrysler’s highly-touted 75-month sales streak actually ended back in 2013 after 40 months. Oops. Now, here’s how the company is sort of fixing it.

Last week, reported that Fiat Chrysler was being investigated by the Securities Exchange Commission and the U.S. Justice Department after two dealers filed a suit against the company for allegedly asking dealers to “create false New Vehicle Delivery Reports”—or, in other words, to claim they sold more vehicles than they really did.

Why would Fiat Chrysler do this? says it had to do with pressure to “preserve FCA’s streak of U.S. monthly year-over-year sales increases, which now stands at 75 months.” This streak could make the company seem healthier than it really is, thus swaying people to invest.

But on Tuesday, Fiat Chrysler denied that it intentionally inflated sales numbers to cheat investors, beginning a statement by clarifying that revenues are unaffected by monthly sales numbers (emphasis mine):

Recent press reports have raised questions about the manner in which FCA US reports vehicle unit sales data on a monthly basis. These reports have in its financial statements .

Fiat Chrysler’s statement then dives deep into its sales-reporting procedure, highlighting its flaws and introducing a new and improved method. Here’s how they framed it.

The Current System Sucks

The two main issues with the company’s current monthly sales-reporting method are processes called “unwinding,” and “not-in-use reserve.”

The company says sales figures are reported by dealers through the “New Vehicle Delivery Report system,” the purpose of which is to lock in the time of purchase so Fiat Chrysler knows when to start the warranty, and also so the automaker knows that it needs to pay the dealer any manufacturer’s incentive that applied to the sale.

Fiat Chrysler says monthly retail unit sales don’t actually affect its financial position anyway, since it’s the initial sale to the dealer that “triggers revenue recognition in FCA US, and not the ultimate sale of the vehicle by a dealer to a retail customer.”

But retail sales do affect Fiat Chrysler’s monthly sales reports, and what’s throwing them off is a process called “unwinding.” That’s what happens when a dealer sells a car, and then—because the customer backed out, or for another reason—puts the car back in its “unsold” inventory. Fiat Chrysler then gets its incentive back, and the warranty period is canceled until someone else buys the car.

The problem is that Fiat Chrysler says it doesn’t consider these “unwindings” in its sales figures, because the company has assumed “that most unwinds are recorded shortly following the time the initial sale is registered in the NVDR system.”

Even though once a car is sold its VIN is blocked from future sales reports (so a car is never counted twice), Fiat Chrysler admits that a dealer can “register the sale in an effort to meet a volume objective (without a specific customer supporting the transaction),” and then unwind it later.

The problem is, if a dealership decides to “sell” 100 cars one month (perhaps to meet a volume objective), and then unwind them the following month, Fiat Chrysler has already logged that as 100 sales the previous month, and does “not...[reflect] either unwinds or the subsequent sales of these vehicles in its sales reporting.”

In other words, after dealers “unwind” sales, Fiat Chrysler doesn’t ever reflect the retracted sales in their books, so it looks like more cars were sold in the month of the fake “sale” than in reality.

That’s a big problem.

The other issue with Fiat Chrysler’s sales reporting process is their method of maintaining a “reserve” of fleet and “other retail sales.” This “reserve” means cars have been shipped, but not yet reported as “sold” in the monthly sales report.

The company seems genuinely baffled by this process, saying keeping cars on “reserve” is “a matter of historical practice (going back many years before 2009 bankruptcy),” but that “the origin of this practice is unclear and is being looked into.”

Fiat Chrysler goes on, saying the “not-in-use reserve” ranges in size each month, and results from a “subjective assessment at the month-end.”

To further show that the company has no clue what the hell is going on with this “not-in-use reserve” policy, the report continues, saying:

...there appears to be no objective methodology for establishing and maintaining such a reserve and thus several plausible values exist for such a reserve. To the extent that the methodology historically used does not yield a unique value, the outcome is inherently arbitrary.

Words like arbitrary, subjective, and unclear, along with phrases like “several plausible values exist” and “being looked into,” just go to show that, frankly, Fiat Chrysler doesn’t know what the hell it’s doing with these sales reports.

On the plus side, the company says that—in general— the reserve is almost always positive, so “in the aggregate, [Fiat Chrysler has] reported fewer sales than the aggregate number of shipped units on a running basis.”

Regardless, Fiat Chrysler says it recognized that this needed to change.

The New Method

So to fix the faults, Fiat Chrysler has devised a new method for monthly sales reporting, one that the company thinks is the “best available estimate of the number of FCA US vehicles sold to end users through the end of a particular month applying a consistent and transparent methodology.”

The new system will account for “unwound transactions,” meaning that, in the case of the dealer who “sold” 100 vehicles one month and unwound them the next, the sales number for the subsequent month will be 100 vehicles lower. And once those now-unwound cars are sold again, those 100 sales will go right back on the monthly sales chart.

As for the “Not-In-Use Reserves” policy, Fiat Chrysler will not just call it a sale as soon as it’s shipped from FCA.

Here’s the “Revised Methodology”:

Dealer reported sales (derived from the NVDR system) will be the sum of All sales recorded by dealers during that month net of all unwound transactions recorded to the end of that month (whether the original sale was recorded in the current month or any prior month); plus All sales of vehicles during that month attributable to past unwinds that had previously been reversed in determining monthly sales (in the current or prior months).Fleet sales will be recorded as sales upon shipment by FCA US of the vehicle to the customer or end user.Other retail sales will either be recorded when the sale is recorded in the NVDR system (for sales by dealers in Puerto Rico and limited sales made through distributors that submit NVDRs) or upon receipt of a similar delivery notification (for vehicles for which NVDRs are not entered such as vehicles for FCA executives and employees).

Chrysler admits that there’s still some slop in the method resulting from things like “unwound transactions that straddle a month end and fleet deliveries, which may be placed into service at various times after shipment and delivery,” but in the end, the company says the new method is better because, above all, it is consistent.

The 75-Month Sales Streak Actually Ended In 2013

Fiat Chrysler applied this new method to all monthly sales reports starting in 2011, and discovered that the “75-month sales streak” actually ended in September of 2013. Here’s the full list of conclusions after applying the revised procedure to 2011-present sales reports:

1. FCA US in March of this year last commented specifically about a “streak” of year-over-year monthly sales improvements since April of 2010. Applying this new methodology, during the periods presented below, year-over-year monthly sales would have declined in September 2013 (-3%), August 2015 (-1%) and May 2016 (-7%). The so called “sales streak” would have stopped in September 2013 (after 40 months) and would have had three additional periods of sequential year-over-year improvements of 22, 8, and 1 month(s).

2. Annual sales volumes under the new methodology for each year in the 2011-2016 period are within approximately 0.7% of the annual unit sales volumes previously reported.

3. The monthly adjustments to previously reported sales as a result of the adjustment to deduct sales later unwound and add back sales attributable to previously unwound sales over the period January 1, 2011 to June 30, 2016 are a mix of positive and negative numbers which did not exceed 0.5% of the reported data in any month. The maximum numerical reduction from previously reported data was 770 units (0.5% of the month’s volume) in May 2015 and the maximum numerical addition to previously reported data was 437 units (0.4%) in September 2014. The total over the 2011 to 2016 period representing unwound transactions previously reported as sold for which vehicles remain in dealer stock at June 30, 2016, is approximately 4,500 vehicles, or 0.06% of the total volume reported over the period (7.7 million cars).

The company pretty much admits that this process for reporting sales sucks, but that it’s not the company’s fault. Fiat Chrysler says there’s no standard reporting process in the U.S., and that other U.S. automakers are probably using a similar method. If that’s the case, why is only Fiat Chrysler being investigated for fraud? The press release doesn’t say.

The company says compiling monthly sales data is much easier in Europe, since automakers report a car as “sold” once it’s been registered. But in U.S., thanks to “diverse recording and reporting practices” in the registration system, car sales numbers have had to be input by dealers.

In the statement, FCA says it “considered simply ceasing to report this sales data on a monthly basis,” and instead just publishing quarterly financial statements. But it decided to continue, because it didn’t want to anger automotive journalists, or look bad to the public.

We understand the sales data are used by some market followers, the automotive press in particular, to opine about the state of the industry and we accept that our decision to suspend monthly reporting would impact those constituencies and .

So basically, Fiat Chrysler is saying: “Our method sucked, so we fixed it. It still sucks, but there’s nothing we can do about it.”

We’ll see if the SEC and the Justice Department buy that explanation.