Hi JT!
How bad can the ST power be? I would think that the single phase ST output, which exhibits voltage changes that vary with the load on the opposite leg, would be pretty easy to fix, and cheap to do, if a fella really wanted to do it. Cheaper than the work you are considering.
Really sweet power is cheap from the sinewave inverters if you really have to have it, and batteries are rocksolid stable. This is even cheaper that "fixing" the ST series.
I have a nice direct drive kubota driven 10 kw industrial lighting plant sold by ingersoll-rand, very nice output, and also a 6-1 driving the cheapest ST 5 I could find and that output is good enough. Not as sweet as the IR set, but ok, and it gives no trouble at all running 'puters and solid state gadgets and pumps and so forth.
Ofentimes I find myself solving problems that don't actually need solutions. I had all sorts of "problems" sighted - most never materialized. Instead the problems were really ordinary - concrete forms that did not quite hold, splices in the first (expensive) ST 5 generator that weren't soldered. Very ordinary stuff.
Provided that the marathon is bolted to a really solid surface - this would take some doing, then the input end of the rotor can be supported - but the marathon shaft is obviously not designed to take the bending moment that an outboard stubshaft and bearing is going to create. This implies that you will need to mimic the presence of an engine with 2 outboard bearings, not one. I make a living building machinery and overhauling power plant machinery - generally turbines - and that's my opinion. I am assuming the marathon is engineered for mounting to an engine flange and not simply adapted.
Much of the "problem" with the ST output is cyclic variation from the prime mover. I expect that the marathon is designed to be driven by at least a 3 cylinder engine. Your twin may produce troublesome and disappointing similar power output qualities.
Best of luck,
P