I haven't read all of that, but the problem at hand seems significantly less complicated.

We're mapping the numbers from 1 to 1000 to distinct numbers up to 8258, and the README claims that we should expect 2.1% of the resulting numbers to be prime. I see no reason for this claim, and as I understand it, the 2.1% comes from pi(1000) / 8258, which seems like nonsense to me.

I don’t remember the 2.1% thing, it could be an error, I don’t know.

I just remember the density of primes was higher: but your explanation accounts for that — well done — because it filters out.