This is where you're getting it wrong, using a magnet would require precise calculations and once the magnetic force drops by a single percent, those calculations would be wrong. If a PMM was made that had a magnet that was supposed to pull a steel ball in its direction at a particular speed so that the ball can fall through a hole in front of the magnet, if the magnetic force of the magnet reduces even a little bit, that speed reduces and the other mechanisms of such a machine would equally suffer for it.