Using Massey Consensus ("bespoke average" ranking for each team) rather than Composite (see below the chart for what those different measurements look like) and converting to percentiles, it boils down to
top 10% of P12 "better" than top 10% of American
next 58% of American "better" than next 58% of P12
bottom 32% of P12 "better" than bottom 32% of American
I do not think this is a surprise to anyone
consensus better reflects differences between ratings -- one spot difference in composite can mean different things depending on where you are in the rankings:
composite 67th Fresno State is two spots above composite 69th North Texas -- but only 1.25 better in consensus
composite 70th Texas State is two spots above 72nd Tulane -- but 3.93 better in consensus
composite 99th Washington State two spots above 101st San Diego State -- but 1.12 better in consensus
there's not a whole lot of pattern to the "value" of one spot in the composite either -- the differences in consensus for 24th to 25th to 26th to 27th are 0.10 , 5.63 , and 0.13
50 composite spots later, 74th to 77th goes 2.37, ZERO , 0.48 (no difference between 75th and 76th, at least to two decimal places)