Incorrect calculation in Picard's CalculateHsMetrics?

[frostzwerg]

I am using Picard's CalculateHsMetrics to create some quality measurements about my whole exome sequencing data set. The output lists absolute numbers of sequenced and aligned bases, as well as bases on-, near- and off-bait (concerning my bait region input file). These are listed at columns "PFUQBASESALIGNED", "ONBAITBASES", "NEARBAITBASES" and "OFFBAITBASES". Additionally, the output contains a percentage of on-bait bases at column "PCTUSABLEBASESONBAIT". Following the description on the Picard website, this should be "The number of aligned, de-duped, on-bait bases out of the PF bases available.". As my reads are already de-duped, I would suggest that the ratio of values under "ONBAITBASES" divided by "PFUQBASESALIGNED" should result in the same value.

Unfortunately, if I compute the ratio "by hand", I get ~0.38 instead of ~0.34 which was computed by Picard.

Can anyone explain the difference? Did I get something wrong or does Picard perform an incorrect computation?

Thanks in advance