"Do the math on displacement. And does anyone know of the link to the Klippel data on the LMS Ultra to verify their Xmax claim?"
i haven't seen the klippel, but ricci's measurements on data-bass suggest a very linear driver.
in the 115db sweep, which produced 110db at 20hz 1m requires that the driver move about 26mm point to point and thd (from all sources not just bl non-linearities) was around 9%.
in the 120db sweep, which produced 114db at 20hz 1m because power compression was setting in requires the driver move about 39mm point to point and thd was around 12%.
so if we take xmax that gives 10% distortion, it would seem to be at least about 30mm, perhaps more.