In this Communication we analyze the relaxation to equilibrium of a biomolecule and its sampling of the equilibrium ensemble, by using simple collective observables. We demonstrate that the customary use of the root mean square positional deviation parameter from the initial structure of a molecular dynamics trajectory, to determine the attainment of a stationary state in a simulation of biomolecules, necessarily leads to an overestimation of the relaxation time; this causes a loss of precious data that otherwise could be used in the calculation of equilibrium properties. A simple and reliable alternative is suggested, by computing the root mean square deviation from several different reference conformations along the trajectory.