Such configurations ALWAYS ignore necessary forces--which are the undoing of the devices.
"Perpetual motion," i.e., machines that just keep running themselves are an EXTENSION of the idea of scenarios where you might potentially get more work from one part of the cycle than another part requires: so it's easier and more direct just to see if you can imagine such a scenario directly--instead of putting it all together into "perpetual motion."
That was the idea with my original buoyancy setup: it seemed like putting the object into one potential energy slide (allowing it to go up the tube from buoyancy) only placed it in ANOTHER potential energy slide (allowing it to fall from gravity). But calculating the energy requirements AT EACH END was, of course, the key factor.
In this case the "changing from narrow to wide" would require more work than can be achieved by the buoyancy harvested, IMO.