I seem to have found a bit of an issue with the Haplotype caller. Looking at variants called with it I've come across a number of small blocks in the genome where the Haplotype caller has called every individual (50 individuals) either RA or RR, which seemed a bit odd considering the population.
Looking at the BAMs and VCFs from SAMtools and the Unified Genotyper these blocks of snps clearly contain all three states as I'd expect RR/RA/AA. Looking at the BAM the reads are of decent quality and have no nearby insertions or deletions to complicate things, and the variants have been called correctly by Samtools and UG.
Any idea what's causing this? Attached is an IGV image showing one of the regions in question, Top VCF is the Haplotype Caller (showing all calls as RA or RR, which is incorrect), Second is UG (showing a mix of RR/RA/AA which is correct). The First BAM shows one of the Animals HC is calling incorrectly as RA for the 5 SNPs shown, while the Second is an Animal that HC is calling RA correctly.
Note these incorrect calls from the HC also passed VQSR. I believe the version of GATK is one of the 2.1 releases.