There are caveats. The lower end Model 3s hit their estimates pretty reliably. The Performance trim usually misses by a pretty big margin for most people. There are exceptions to every rule, but I owned a P3D and hung out with other P3D owners as well, and this was a commonly shared experience.
IIRC there was some youtube video a year or two back that tested a few variants on a track using what they claimed to be a replica of the EPA highway test, and the cars came out almost dead on.
There have been some independent tests and IIRC the consensus is that the standard range Model 3 was rated pretty accurately, and the Performance trim was rated something like 25-30 miles too high compared to reality.
But again, you're using the words "on the highway" to reflect what I assume to be your driving habits in rural open road conditions and not the test that was actually performed to produce the numbers in the product specs. And that was my point. It's not about advertising fidelity at all, it's about regulation and its mismatch with the expectations of a modern EV driver.
The Tesla numbers are correct and verifiable, actually. The problem is that the EPA "highway" test is designed to capture the "average highway mile" driven by a US vehicle (which makes sense, it was designed to inform purchasers of aggregate fuel use). Here's the driving schedule:
https://www.epa.gov/sites/production/files/2015-10/hwfetdds....
Note that it never goes above 60mph, and spends most of its time between 40 and 50. This isn't wrong, but it corresponds to an urban highway commute (again, which is pretty much the "average highway mile").
The problem is that this does NOT match what typical consumers think of as "highway driving", meaning blasting 75mph+ down an open road. So people get in their Model 3's and measure their consumption at those speeds and it doesn't match.
And it's true that other EV manufacturers (Ford, especially -- the Mach E actually does somewhat better than sticker consumption on the highway) chose not to run the full test and used other methods that put a higher number on their stickers. And in context that's probably more informative.
But nothing was "rigged". In fact Teslas in stop-and-go urban commute traffic really do get close to their sticker mileage. It's just that no one cares because one's battery doesn't run out on an urban commute before you get home to charge it. Ironically, EV owners in this case care LESS about efficiency than gas owners.