I suspect there is much more going on "under the hood" in the execution of mega voices.
I have never really dissected a song file created by a Tyros, to reverse engineer so to speak, but I wouldn't mind betting there is also some serious sysex being generated at the same time. Sysex that would ONLY be programmed into the Tyros style data.
There are limited mega voices on the Motif ES and when I have tried to play Yamaha styles (using the OMB software) and trying to utilise these mega voices the style always "glitches" and through a process of elimination I have concluded there is more happening in the CASM section and with the mega voice than is apparent.
If someone can record a simple SMF using a mega voice style, it only needs to be 8 bars, and send it to me I would be interested in attempting to investigate just what is happening. I had a tyros 1, but never thought of actually recording a style to an SMF, but sadly I no longer have it.