On improving the Shortest Length Confidence interval for the Generalized Variance