It's clear that drop-off varies a lot: ±30% by state, and even more by county. I'm sure if you looked at the vote data from every tabulator in every county in every state, you'll find more than one "squint and you'll kind of see it" pattern.
I'll need more evidence before I suspect anything bad happened to the vote. (IMO voter suppression efforts are potentially election-impacting, but vote manipulation less likely.)