This is exactly the sort of thought-out, nuanced feedback that brings a smile to my face!
Currently, I'm trying to not do too much balancing while the game is still in flux each update. But I will share with you how I do do balance: I have the AI play against itself for hours on end in simulated battles, recording win-rates, and adjusting numbers accordingly. Those simulated battles usually have fewer units than the big week7+ battles that we've all tried. So as you suggest, they probably work a lot differently.
I think, if nothing else, with your feedback, I should go back and make sure to update the simulation to also use the reserve system and have more realistic numbers of units. I have your hypothesis below to research - and if you're right - down goes the nerf-hammer :D