Because there was a ton of them, and even more with >29% (which I didn't submit). The comment there says it all (they are a dataset for anyone who wants to learn CHG in practice).
The real search target were
twins. (and the
sieve was for twins, so whatever was found is far from a complete list primes/PRPs)
Kamada has a
systematic (x,y) pair list up to a certain limit.