Right you are, Tim. I did a bunch of mathematical modeling during my career, both good and bad. Modeling is a powerful tool for designing and understanding things, but the worst mistake you can ever make is to loose track of the fact that it is always only a model.
Just like building a physical model of a generator (see how I'm trying to pull the thread back on track?) You need to decide which parts you want to model accurately and which parts you don't care about, which parts will just naturally work and which parts you need to have to change because they're in a different environment...
as as my late father put it "you can't scale nature". With so many square and cube laws at play here especially with Electro - Magnetics in these small scales some things need to be either overscale, underscale or done in a different way in order to get these things to work at all.
At these small sizes phenomena which in full size are largely insignificant and can be ignored start to have much more pronounced effects.
This tends to make Computer Modelling, as good as it is, less reliable. The best way is a physical model, and experimentation, just as the OP has done, (I'm trying to pull the thread back too !)
Such is his dedication, that he has wound several sets of tiny coils until he got the balance right and nearly went "round the bend" (ouch, bad pun) doing it.
At the end of all this hard work and experimentation he has produced a working, practical and I think beautifully period looking alternator.
His design is not quite as unconventional as you might think "coreless" machines do exist and they are widely used especially in applications where weight is a concern. (like drone motors)