I was pretty sure that I needed to wait a few more days, and to also let the rubidium standard stabilize for a longer time too. I was just pleased at how low the difference was already. What is the math to figure out the frequency difference? I am too lazy right now to figure it out.

Well, I am certainly new to this area and not an expert, but I think there shouldn't be a frequency difference once it's locked. I would expect it would vary by some fraction of a full period, but the GPS-driven adjustments should bring it back. That seems to be what I am seeing, but I am not viewing it over a very long time.

If there is a frequency difference, it will "slip" N cycles over some period of time, T relative to the reference. The frequency difference is N/T. So for example, if you hooked source 1 up to a counter and incremented that counter every cycle, and hook source 2 up to another counter and increment it every cycle, and after 1000 seconds of observation the two counters differed by 2 counts, then the frequency difference would be 2/1000 = 0.002 Hz.

(Adding fractional cycles)

If it's slipping a fraction of a cycle, just plug that in - so 5ns is 0.05 of the 100ns period of a 10 MHz source, so 5ns in 80 seconds is 0.05 / 80 = 0.000625 Hz