If you mark one part of a joint from another, what's the error associated with that procedure? Isn't this what determines the accuracy of a joint rather than the ability to measure anything? (I'm assuming here that you cut the joint by hand.)
I started wondering about tolerances when I started trying to face joint wood. If I aimed for "perfect", meaning that I couldn't detect any deviation from my Starrett, then I'd still be jointing. And I concluded that seeking that level of accuracy doesn't make sense because of the changeable nature of wood. (I plane the other side and it bends 0.1" anyway.) So if perfect isn't the goal...then what is? I picked 0.004", but I don't know if that's the right answer.