It is a trade off of game flow (less measurements) vs a small inaccuracy where it doesn't matter.
Take a notice next game. A punt always gets spotted on a major yardline where possible. A punt inside the 10 of course would not be possible. Rounding from the 2 to the 5 would have a huge impact. Or if it lands and comes to rest directly between two major lines.
ex: The offensive team starts from the standard 20 yard line. Then during the first play, a pass is made, and "block in the back" is called against the defensive team. The penalty would be 10 yards, and would put the offensive team at the 30 yard line for the start of the play.
There is also the flaw in that these measurements are reported by humans so we can't tell the source of any bias that might exist in the numbers. Similar to the glasses of water question in the lede, I think it is more likely that any bias comes on the reporting end rather than in the actual data. The refs don't spot the ball always on yard line so they might spot the ball at 38.4 yards. The person reporting that value is judging it by eye, so they will generally report that as on the 38 but will occasionally report it as being on the 39. Does the chance they report a ball that is at 39.4 yards as being on the 40 yard line increase in comparison?
Couple notes - 1. If 'normal' plays have a tendency towards a 10/20/30 yard line, what about plays where the ball placement is challenged? Don't know if your dataset has this, but under your framework, we'd expect to see a more even distribution of placements because of increased attention, though probably still with spikes at the 1 yard line and goalline. 2. Even though it's probably intuitive, the unimodal distribution of ball placements was cool to see.
