> iirc ~75% of propellant mass on Starship is LOX
I remember something similar. The stoichiometric mix is 80% oxygen, but the actual mix is fuel rich, because CH4 is a light molecule, and the end result is a higher specific impulse.
Still, one could imagine that you could run the rocket in oxygen rich mode; you give up some specific impulse, but you might need to carry less CH4 with you for the return trip.
All that said, I just run the numbers, and I had a huge surprise: if I didn't make any mistakes, the Starship has enough fuel for the return trip, without any need for fancy refueling in Mars orbit. According to the wikipedia delta-v map [1], the delta-v needed to go from GST orbit to the Mars transfer orbit is only 1.16 km/s (yes, not a mistake). Then from there to low Mars orbit it's a further 2.1 km/s, but aerobraking is possible. Also out of the 1.16 km/s from Mars transfer to GST, 0.77 km/s can be done with aerobraking on the return trip. But let's be conservative and ignore all the assistance from aerobraking (although the Starship was designed specifically with that in mind). We end up with a total of 3.26 km/s for one way, and 6.52 km/s for both ways.
The exhaust velocity of the Starship is 3.56 km/s [2], which leads to a ratio of delta-v to exhaust velocity of 1.83, the exponential of which is 6.24. So you can have a ratio of initial mass to final mass of 6.24. The ratio of the Starship gross mass to dry mass is 1300/100 = 13, so more than twice the minimum. Of course, the delta-v map assumes the most economical routes, and for human travel you might need to splurge a bit, to get up there faster. Still, 13 vs 6.24 seems like a good margin to me.
[1] https://upload.wikimedia.org/wikipedia/commons/9/93/Solar_sy...
[2] https://en.wikipedia.org/wiki/SpaceX_Starship_(spacecraft)