Not an anomaly per se, I think, it's just that the machine rounds up, instead of rounding to the nearest.

Did you get that result through repeated trials, or only one pairing? The original poster claims that the decimal-children would be likely to, in this case, be a 12 OR a 13, without a greater likelihood of being one or the other. (With drift into the 10, 11, 14, 15 range)