South Minneapolis is pretty much unpopulated other than parks/schools/churches:
Same in a good chunk of Memphis:
and some major suburbs of Atlanta:
It appears to be a usability/functionality tradeoff. Apple Maps isn't perfect in privacy nor accuracy/completeness, but OSM is useless for many people.
I once picked 100 random coordinates on land (by picking random coordinates and filtering out those that landed in water). OSM is by far the best. It's not even a competition. I still have to do a writeup of that some day, posting just the stats and methodology (scientific paper style) won't interest most people, and I forgot to capture screenshots, so I basically have to redo it completely.
It's almost certainly different if you select random points biased by population centres, but at least know that "not remotely usable" is a very local issue (see the sibling comments about the Netherlands) and that it depends on your use case.