"Exact" deviations in Wasserstein distance for empirical and occupation measures