That's actually almost certainly a statistical fallacy due to not knowing Bayesian statistics.
The number of post-cap Cup winners is 20, not a big sample. If you go through dozens of advanced stats, you actually expect that some of them will give you better retroactive predictivity than e.g. standings due to luck and small numbers counts. That's a form of p-hacking actually, and similar to the look elsewhere effect. The more questions you ask, the more likely you are to get a spurious result just due to luck
This leaves aside the fact that playoff hockey is quite distinct from regular season hockey. There are fewer power plays, the backup goalies don't see ice time, there are no bad teams to beat up on, no 3on3 overtime, etc. It makes sense that a 5on5 statistic like corsi would be more reliable in the playoffs, but not because of voodo, rather it's because of the rule changes.