You can get big differences with and without autostop if there are more than just a few stops en route. It's not unusual for my 3hr ride to have 9 minutes of stoppage time, all at stoplights, and I'm not riding in urban areas. If you gain confidence that you can hold 220 watts avg power for 3 hours, but in reality, your autostop is giving you 10-15 minutes of downtime not included, you might get a rude awakening on race day when you no longer have that stoppage time.

You can see how huge something like an autostop at a light would be in a 20min FTP test. In parallel, how hard would you have to train to get that 20sec of stoppage time in a running 5k of 20 minutes? Probably a lot.

When I used to train in urban LA, it was crucial to not use autostop, as I'd get these massively inflated times/paces because I could run fast for the 4-5 blocks between stoplights (without even trying to go too fast - you're just rested from the stop) and it was not good to use those to gauge racing.

In my experience though, stoppage time doesn't seem to affect NP that much. (It affects average power much more.) I know the calculation for NP, but at least how it plays out in real life for me, is that it's an estimate of what kind of potential avg power you must have to be able to DO the workout you did. So if you only ride 20 minutes total but throw down only two 3 minute intervals of 350 watts (meaning you rest for the other 14 minutes coasting), you'll get a really high NP, whereas a 3 hour ride at steady aerobic pace that would wipe you out much more (much higher TSS) will have a much lower NP.