Yeah, I was wondering about that. I found an explanation of "fractional medians" at http://www.stata.com/statalist/archive/2009-01/msg00662.html - that might be how they came up with the number. Not sure how relevant it is though because the explanation describes it as a way to get a more precise measurement when the data is artificially constrained to, for example, whole numbers...while still retaining some of the benefits of using the standard median.