The problem with the conversion to Yamahan is that when a sample is using by many sound presets, they are duplicated in the Yamaha. I don't know if many motif sounds can share the same audio sample
But that's not the reason here. You are referring to duplicate waveforms in the flash board (which are avoidable, the Motif/MoXF can use the same waveform for multiple voices, you just have to take care yourself that you don't copy duplicates onto the flash if you load several voices one by one). But here the E-MU package already has hundreds of MB when opening it in .x3v-format in the Melas waveform editor on PC. The waveform editor doesn't duplicate any samples, obviously it is the .x3 format itself which inflates the data size.