The heuristic of A*/A-star search and its efficiency

Question

I have a feeling that the more the heuristic underestimates the real cost, the more inefficient (expanding more nodes) it is for A* search to find a optimal solution for path fi