Towards a Deeper Understanding of Concept Bottleneck Models Through End-to-End Explanation