Emily,
I don't understand your question. It is not unusual for the group median to equal a value in the data. Where it gets somewhat messier is in that the definition of "median" can vary and that can impact the estimations. Note that even though these p-values are called "exact" they are not the brute-force computations that we learned in graduate school (those approaches rapidly become prohibitive when the sample size exceeds 10).
Perhaps you could read the documentation referenced in
http://support.sas.com/documentation/cdl/en/statug/63962/HTML/default/viewer.htm#statug_npar1way_a00...
and reframe it.
Doc Muhlbaier
Duke
