Missed in the initial sieve yes, there were quite a few missed during intial sieving for one reason or another.

Client, user, not-submitted etc... This was the driving force behind the 991<n<50m effort. I think you will find that we missed quite a few hundred but not thousands.

Second you probably don't have all the data for p<25T, after some time after p>25T was sieved an agreement was meet to cut the results file at 25T.

What's important is if you find any factors <50T (currently) that factor some k/n pair in the dat, either mikes updated 2.7m<n<20M or our 991<n<50m. You won't find any if you do there are serious problems.

I've been looking at your data a little closer not sure what to make of it yet. What is your goal, just info or is it geared towards a particular future aspect of the project. I can see it's use in P-1 factoring but not nessarily sieve. It would shed some light on factoring bounds what to expect etc if you made the data specific to 900T<p<2^50 and (8M<n<15M ???). But for sieve it's always most efficient to start with the smallest un-sieved p.