This depends a lot on the power switch element and the
DC motor attributes. Your armature inductance will be one
concern, you need ripple current to be sensible (maybe
10% of full load amps) and you can get that from worst
case max line voltage and motor inductance (though I'd
guess maybe this is modal too, what's the worst case
minimum inductance found at?).
Then you have the tradeoff between conduction losses
and switching losses which for any DC load and device
type (IGBT, MOSFET, bipolar, GaN) have a "sweet spot".
IGBTs like to run slow but are the winner at very high
bus voltage and current (less so at low voltage where
their saturation voltage eats up efficiency). MOSFETs
as you indicate just have a fairly fixed Qgg per cycle,
and a fairly fixed conduction loss per cycle (Iavg*V*ton)
and you can play with that.
I recommend making an -in-efficiency oriented spreadsheet
where each of the loss terms is individually figured. This
is a useful design tool, unlike rollup efficiency numbers.