Revisiting the false positive rate in detecting recent positive selection