I will agree with you on the last point.So they use strength of sched, off and def avg, so if you don't use either of them are they only basing their ranking on strength of sched? The eyeball test tells me more than any computer and nowadays you can watch dam near any team you want on livestream. Even these poll rankings are a joke because the only thing they look at is record and do no research at all like who they played.
As far as how the rankings are done. They look at every individual game, and what the spread was and then optimize the score differences across all current and previous games which essentially generates a rank. So in essence if there are two common opponents, teams that beat the common opponents by more will result in a higher relative rank. Do that for every game throughout the entire season and it gets pretty accurate. Anomalies should work themselves out. At least this method is using multiple variables to determine relative strength rather than just w/l.
There is no way you can look at a team and all the individual matchups within a game to know what the net effect of each team is - too many variables.
Agree score is not necessarily 100% foolproof, but it is better than just looking at record, regardless of opponent. And I agree that individual matchups matter, but that only works in a head-to-head comparison, you can't develop a complete ranking system that way.