It turns out it was due to rotational weight!
I noticed that MagLev cubes don't have trouble with M slices, and I got to thinking...
There are 20 magnets in an outer later, in an outer radius...
But only 8 magnets in an M slice!
If it was due to magnetic force, people would be UNDERshooting, and not OVERshooting.
So it's not due to magnetic attraction, it's due to torque and rotational weight effects!
By balancing the M slice by gluing washers to the caps, you achieve weight balance in the M slices, and you can nail your H perm, Z perm, O perm, U perms, ezpz.
Since the washers are in the center, you don't feel it when you do outer layer moves, ONLY when you do slice moves. The cube as a whole is heavier, but since the weight is near the center, the torque required to make the turn on the outer layer isn't affected much, but it's greatly increased with the M slice, which balances it with the outer layers!
EDIT: Uh oh, if you pause at 3 seconds, in the background you can see the scheme, formula, and chemistry behind all three steps of Project Boron-22. Good thing I don't shoot in hi-res HD LOL