What files would be the most important for troubleshooting if my results don't match my expectations?

The *_primers.bam file contains a superset of reads used for variant calling, by checking this file using samtools mpileup (or other similar tools) you can determine if the variant of interest was present in the dataset to begin with. In case the variant was present in the BAM file but it did not show up in the final results it is likely due to multiple variants being present in the activity window of the target and GATK failing to phase them accurately. If this is happening due to germline variants the recommendation is to update the reference genome such that the reference genome matches the germline variant.