This isn't true though. Middling P6 coaches don't have teams that have higher KP's than mid-major guys just as a matter of course. Wofford, Lipscomb, Belmont, Liberty, Saint Mary's, Murray State, Furman, Toledo, and New Mexico State are all in the top 60. The ACC has like six teams worse than 60th.
The point is not that there is a predictive correlation between a higher computer ranking and being a "good coach" but rather this is a better relative method than simply looking at wins. The computer rankings take into account all the games that are played (they don't emphasize one big win in the NCAAT) and they are forward looking (how good a team is moving forward, rather than an indication of where a team should be ranked based on the past).
Further, KP adjusts for strength of schedule in its efficiency based on both offensive and defensive points per possession of opponent. The entire point of these systems are to remove the apples and oranges comparison of "is UMBC as good as Belmont?" and normalizes information across the board.
I'm yet to hear a legitimate criticism of KP or other advanced stats that are substantively rooted (addressing methodology) in comparison to those criticizing outcomes ("well Penn State can't be a top 40 team!") Well why not, what is it that KP is specifically calculating incorrectly to get this result?