I'm assuming you mean MOXF. If that's correct, MOXF doesn't have an FM engine or expansion card for FM. Therefore, what's done with this conversion cannot be done on MOXF. You could sample a DX7 and load those samples into flash - but DX7/FM programs are unique in how they respond to velocity (dynamic response range) so the _behavior_ of FM would not be easily (or even possibly) sampled.
The DX7 stores its sounds as a finite series of numbers that takes up hardly any space at all. Compared to sampled based sounds that can take quite up a bit of space for the sample data. The meaning/organization of this finite number series which represents a sound is also a part of MODX (and Montage) so these sounds just "pop in" to MODX without a hitch. Conversely, there is no such thing in MOXF so you cannot "pop in" these sounds to MOXF.
Note if you go further back in time - earlier Motifs had expansion cards that could do, if you had the right card, FM. MOXF/MOX is beyond that time when expansion cards were dropped.