I think this method is a bit of a fool’s game. Too many variables (bit width, bit deflection, trying to accurately measure kerfs cut into wood using a set of calipers, it all stacks to problems).
If you were able to get accurate parts in one area of your machine previously by using the indicator method, but larger parts were inaccurate, then I’d suggesting finding a way to mount a set of digital calipers to your spoil board and running the same sort of test over larger areas.
Mr. Adams indicated that there can be uneven belt stretch and maybe that is what you’re dealing with. Maybe your left and right belts on the y-axis aren’t closely matched or stretched differently.
A 6" set if calipers would be a good start. You can get 12" calipers on Amazon for not too much money.