Cluster without fluster: The effect of correlated outcomes on inference in randomized clinical trials