We are running an exome sequencing project where we have sequenced samples on two different lanes. Aligning with bwa, we assign identical ID and SM tags, yet different PU tag for these files. This should be in line with your general recommendations, keeping the lane information available for recalibration purposes. We have then fed both files from the same sample into the base recalibration step to create a common, sample-level bam, also in accordance with recommendations previously posted on the forum. However, when calling variants we get VCFs where UnifiedGenotyper has treated the different lanes as separate samples. What are we doing wrong? Is this approach not possible after all, so an identical read group is required for each sample?