On the empirical variance of a fractionator estimate