A note on the asymptotic variance at optimal levels of a bias-corrected Hill estimator