As I understand it, Abra is a tool that takes mapped reads and realigns them around indels, because the reads originating over the indel were not aligned correctly. Locally realigning over suspected indels can help improve alignments, if the mapping (approximate location) is correct. However, it will not place reads on indels if they were mapped somewhere else in the first place, because the mapper could not tolerate long indels and thus did not consider that a valid mapping location. So mapping with a program that cannot tolerate long indels, and subsequently realigning with a tool that can, will generally yield inferior results to mapping and aligning with a single tool that does tolerate long indels... because in the second case, there will be a lower rate of reads being mismapped or left unmapped.

If the goal of the thread is solely to discuss Abra, then my comment is irrelevant. But if the goal is to find tools that do a good job of aligning reads around indels, then it is relevant.