Comparing incomparable seasons
Two teams may share almost no opponents, so any comparison between them is inferential. Whatever method is used, it is estimating rather than observing. The output is a numbered list, which looks authoritative and hides how much interpolation went into it.
Opportunity dressed as merit
Resume-based criteria reward playing strong opponents, which depends on scheduling and conference membership as much as on how good a team is. A very good team with limited access to quality opposition is penalised for circumstances it does not control.
So treat the numbers loosely
The practical consequence is that adjacent seeds are effectively indistinguishable, and arguments about being placed a line lower are usually arguments about noise. Broad tiers are defensible. Precision is theatre.