So, simplifying, Runeson claims that the reason that the static aperture viewing of the Ames room leads to the perception of a rectangular room is that, in the real world, the trapezoidal room is so unlikely to exist. The trapezoidal case is, as Dretske and those armchair philosophers might call it, not a "relevant alternative".
Ok. Now, this solution would not seem to me to help with the amodal completion of the pac-man to a circle. We univocally see a circle even though the pac-man is not a lawfully improbably entity. The laws of physics/geometry do not make it highly unlikely that the world will contain a pac-man. Indeed, free dynamic viewing reveals it to be a pac-man. (And, really, one gets a lot of different things to be amodally completed to a circle, e.g. a tear drop shaped thing with the pointy end occluded. Is the circle really more probable than any of these? Or maybe is it more probable than all of the others put together?)
[Incidentally, the dynamic free viewing of the pac-man doesn't make the illusion go away. This is apparently unlike what happens with dynamic viewing of the Ames Room. The Ames Room illusion apparently fades considerably with free movement away from the aperture.]