yeah, I see what you mean. Only reason I suggested a screen shot is I find it easier to visualize seeing the grids. I would guess since solves look at optimum play vs. optimum play, that 33% bet sizing is going to yield the best results vs. an opponent that is playing optimally (and sub-optimal for that matter). So it will vary the frequency it bets with a hand like AT on A96 or whatever board. In this case, the T blocks some OESD and gutters and it's going to take that into consideration. I would expect the solver to use a very balanced bet/check strategy in this situation. It would make sense to bet AT for value on A96, especially if you knew your opponent had 87, since you block 25% of his draw outs. On the same token, it would make sense to check AT on A96 if you knew your opponent had 87 or T8 or T7 or something like that because you can extract a bet from your opponent when they bet their weak draws or weaker pairs.

I would also guess since the boards you described are somewhat coordinated and connected, and this will have an effect on frequencies/sizes. I would assume that it will bet wetter boards more often and also for larger sizes and bet drier boards less often and for smaller sizes.