For what it's worth, I will present my 2 cents on this subject.
I've often found my train of thought following along the same lines as what you've presented here.
High Voltage versus Low Voltage versus a given wattage of output(load)
This all has to do with resistence. Resistence is the root of all losses of efficiency.
Resistence in the wire, resistence in the fet's that do the switching, resistence in the Lipo Pack itself, and resistence in the motor winding.
Another thing that gets tossed into the equation is voltage drop.
For a given length of wire, you will see a measureable amount of voltage drop for a given load. The greater the load, and the smaller the wire, the greater the voltage drop you will see.
We try to make up for this by using larger wire that has less resistence. There's that word again.
If resistence makes for more voltage drop with the greater amp load, then it makes sense, by bumping up the overall voltage, we effectively drop the amp load and minimize voltage drop.
By dropping the amp load, we lose less. We lessen voltage drop and we lessen wasted energy spent on heat, because of resistive losses.
The cooler we can make something run, the more efficient it becomes.
But we push the limits. The limits are pushed with the low voltage setups AND the high voltage setups. I think this makes it harder to measure the differences.
I seen a good example with my lithium ion trolling motor setup. By eliminating a 7 foot 10Ga. length of wire, the motor would put out an extra 75 watts of power. This was the only change that was made.
Efficiency was also improved, because that 7 foot length of 10Ga. was noticably warm to the touch. Even though, it was "only" carrying 50 amps.