It depends from case to case. If the heat treatment parameters are properly controlled, then the microstructure should be representative. You can say dilatometry gives you the ideal result. If you are in doubt, it's better to take through-thickness microstructure from top to bottom surface at a reasonable interval.

Dilatometry experiments performed on standard samples can give an indication what will happen in a big casting block. This indication should be taken into account with a "limited responsibility". The main reason for a difference between the "ideal" (standard) and the real (big block) is micrtostructural changes occurring as a results of different cooling rates in these two experiments.