The current coverage calculation doesn't look strict enough. We're looking at coverage (roughly) in terms of (last_base_aligned_with_ME - first_base_aligned_with_ME) / insertion_length (and also discounting polyA/TSD, although that's not the main issue.) So we get something like this with ~92% coverage but the real number is clearly lower:
Try a strict % bases covered and see what it does to the results.
The current coverage calculation doesn't look strict enough. We're looking at coverage (roughly) in terms of (last_base_aligned_with_ME - first_base_aligned_with_ME) / insertion_length (and also discounting polyA/TSD, although that's not the main issue.) So we get something like this with ~92% coverage but the real number is clearly lower:
Try a strict % bases covered and see what it does to the results.