Re-Visiting Explainable AI Evaluation Metrics to Identify The Most Informative Features