The logical falicy would be if they ONLY measured successful bunts.
Not every bunt attempt would be successful, which needs to be accounted for. Failed bunt attempts are a possibility.
Just like not bunting will sometimes be successful. That needs to be accounted for too. And will fail, sometimes.
But bunting, or attempting to bunt and failing, led to more wins than not bunting, even accounting for the times one if the next hitters gets a hit and drives in the winning win. Which is also a possibility.
I think you have the logical falicy backwards.