Actually there are 78 primes for 19000 <= x < 20000. I miscounted above.
All x < 20000 has now been doublechecked. No missing primes/PRPs.
My tactic for testing 20000 <= x < 30000 is a little different due to speed of sieving. I am testing all x in that range at once, but subdividing by y as PRP tests for smaller y are faster. I am testing for y <= 1000 while sieving for y > 1000 continues.
