I wouldn't call running samtools fixmate on a bam as validation, since that's not its task.

Just a guess, but are you name-sorting the bam file before running fixmate? This is required, per the documentation, presumably so that reads are adjacent to their mates in the file.

I don't know what will happen if you run it without name-sorting the bam, but it may produce low paired estimates like you see.

A more natural way to actually validate a bam file is the Picard tool ValidateSamFile.

Thanks for the question - good clarification! When we run samtools fixmate on a sample BAM, it reports that only a miniscule (2.48%) of the reads were properly paired. This seems very strange. Is there any reason why samtools fixmate might give misleading information for Complete Genomics BAMs?

Thanks!