Kenpom across years: a fools errand
I pointed it out earlier but comparing kenpom teams against years on an objective scale is invalid.
This is why in chess, for instance, you can't compare ELO across populations which don't regularly play eachother...which is why computer ELO != lichess ELO != chess.com ELO != FIDE ELO.
Kenpom is a strong analytic tool, but it's not good enough to compare teams from populations which don't play each-other. A team's adjusted efficiency is based on their expected performance against an average team. There is no guarantee the average team across years is the same. We can say with pretty high confidence that they're close, but there's no way to definitively know. Further, how far the outliers move from the mean depends a lot on the overall distribution of strength of the teams. If a lot of teams are clustered tightly around the mean, then it might artificially inflate outliers, whereas in a flatter distribution, it might suppress them.
So the best we can say is these teams would be among the best in the kenpom era. What we can't say is these would be the best KP teams were they to play in any season since 2002.
It's fun to think about how these teams would rank among other great teams in this century, and maybe they are the top 3...but KP EM is not the right tool to make that case.