Statisticians very often control for proxies when the actual causal variable can’t be measured; for instance, party affiliation might be used as a proxy for political beliefs. Because Z isn’t a perfect measure of M, some of the influence of X on Y might “leak through” if you control for Z.
