Ah, ok, my VI was inadvertently measuring the resolution of the Windows clock, which is why there's a pattern. Using the tick count, I get accuracy within 1 ms, as the documentation claims for Windows based machines.
Thank you all for your help resolving this issue!