Is stationary distribution of markov chain better approximation than class distribution in sample set?

Question

Given a sequence of states in a system, we can compute the transition probability matrix, and using stationary distribution of this transition probability matrix we can esti