I've found down interesting numbers around high-leverage situations that explains a lot of the disconnect between down-neutral stats (CPOE, passing grades, Y/A, etc) and EPA when it comes to Patrick Mahomes and Russell Wilson

In short: Patrick Mahomes has been really, really good at converting 3rd/4th downs and has been doing it consistently his whole career. Mahomes has converted 3rd/4th downs at 10%+ over expectation (accounting for distance yard line) each of his three seasons

Wilson has been roughly average. So from 2018 to now, Mahomes has converted, through pass or scramble, 45 more 3rd/4th downs than expected and Wilson is at -1

It makes sense that stats treating each play equally can see Wilson as better than Mahomes but Mahomes still have higher efficiency in EPA. He's generating a ton of EPA in the high leverage situations to make up for poorer performance on early downs

Mahomes' 3rd/4th down performance is likely to regress this year (he's at almost 20% over expectation), but that regression is to something like 10%, not 0%

Most likely scenario is that Mahomes starts playing better on early downs while getting slightly worse on late downs, which means more like 2018-2019 Mahomes