While I agree with your final conclusions, I think your approach can be questioned. First, the Newtonian laws you reference refer to inelastic collisions; air is very much an elastic medium and its compressibility does affect the energy transfer (albeit considerably less than it might under other conditions). Second, you seem to infer that a driver's distortion-limited X(max) is defined by its suspension. I don't think this is the case nearly as often as you might think. Rather, the distortion-limited X(max) is defined by the driver's motor geometry; the suspension, hopefully, serves primarily to keep everything in line. And even in the lowest-compliance designs, where the suspension supplies a significant portion of the damping, it can behave in a remarkably linear fashion if properly implemented.

I agree that the advantage enjoyed by the larger drivers results from the shorter stroke necessary for a given pressurization, but I'd argue that it derives primarily from motor geometry. The increased X(max) required for the smaller driver to yield an equivalent pressurization would require, assuming equal gap heights, a longer voice coil. This creates two problems. First, the lower gap height/X(max) ratio of the smaller, longer-throw driver means that the BL will drop off much sooner and more quickly than it would with a larger gap height/X(max) ratio. And it will take the motor's ability to control the cone's motion with it; increasing distortion at a given level. Second, the longer voice coil contains more wire and, presumably, more turns than its shorter counterpart. And since this greater number of turns is moving through a longer stroke, thereby breaking more flux lines, the back-EMF increases considerably, further diminishing the driver's output for a given (open-circuit) input.

Or I could be completely wrong...