A reader who may or may not want to take credit Comrade PhysioProf has pointed me to an interesting factoid on the NIAID site. Page down to the section on "Comparison of NIH's Old and New Peer Review Processes" and you will find:
[Old way] Percentiles range from 0.1 (best) to 99.5 (worst).
[New Way]Percentiles range from 1 to 99 in whole numbers. Rounding is always up, e.g., 12.1 percentile becomes 13.
So not only do we have scores clustering around the even-integer increments allowed reviewers, we have additional clustering of applications based on percentile rounding.
As far as a Program Officer is concerned, your 9.1 %ile application is the same as the next person's 10.0%ile application.
Man, they really were serious about intentionally generating a lot of tied scores, weren't they?