There is an argument that good science sets the parameters in advance and follows through with it. Fiddling with your methodology after the fact is one way to create results out of nothing, because you can find the things that was, by random happenstance, an outlier variable.
Rolain’s methodology was to perform mouth swabs. He ignored people who he couldn’t swab - like people who died.
That he stuck to it is, in the previous sense, good. But it is, obviously, a bad methodology. Sticking to a bad methodology - particularly after people have pointed it out - is also bad science.
If his new study uses a better methodology then that would be good. If he stuck with the old one, then I would be pretty hesitant to accept his numbers until I knew more about what data went missing midway through.