Revisiting the Maximum Likelihood Estimation of a Positive Extreme Value Index