Quoting Wikipedia as evidence, nice.

Actually in the recent Model S test prior to this one, Tesla discovered through examination of the data logs of the car that the EPA had carelessly left the key in the car all night while it was parked during testing that spanned multiple days, leading to an artificially low number as the car was unable to sleep at the level it normally should be able to when parked.

This whole episode happened because the EPA was doing the testing… I don't know where that leaves your anecdote. It's just false, I think. I love Wikipedia and even have a monthly donation to them, but your link is not reflecting what happened with Tesla here.

OK, fair point. Allow me to quote from the EPA directly. [1]

"Manufacturers test their own vehicles—usually pre-production prototypes—and report the results to EPA. EPA reviews the results and confirms about 15%–20% of them through their own tests at the National Vehicles and Fuel Emissions Laboratory."

Wait a minute...

[1] https://www.fueleconomy.gov/feg/how_tested.shtml

And I'm not sure how an indication that a Tesla has been tested (and retested) invalidates a claim that 15-20% of all cars are tested. There have been dozens of Tesla models over the years, one would expect at least a handful to have been verified.

Which by the way would have totally sufficed to invalidate the original argument; the ranges are largely self reported, but they do get checked. The process ensures that the ratings broadly reflect EPA's testing methodology, otherwise you'd see the cars that do get tested having abnormally low range at best, and fines and other action taken by the agency at worst.

A Wikipedia link might be a crappy form of "proof", but as near as I can tell the parent post just made up how he or she thought the process worked. If someone says Caesar lived in the 9th century, a link to the relevant page actually is a valid response. If that person comes back with some source material that says otherwise, then you'll need to try harder.