It looks like something is degrading your results. The 53132A has a resolution of 150 ps. This should give you a one second result of 1.5e-10 but you're seeing more like 6e-10 and it's consistent through all the tests. If you haven't done so, check the noise floor of your 53132A by connecting a signal to the Ch1 input through a T connector and a cable to Ch2. Set Ch1 to 1 Mohm and Ch2 to 50 ohm. Measure the time interval from Ch1 to Ch2. Make sure that the cable doesn't move or change temperature during the test. If that test is good, now you're into the black arts of AlDev measurements. Ground loops, power supply coupling, injection locking, external and internal interference, stray sunbeams or other temperature changes, etc. After all, we're mucking about in the PPT area. Be sure to think positive thoughts! Try to stay at least 2 meters away from the equipment during the measurements. No, I'm not kidding! (about the 2 meters - not sure about the positive thoughts, but it can't hurt
)
The Rb-GPSDO tests are about what's expected. Looks like there was some warmup and stabilization between the two tests. GPS performance can be broadly defined as a line that goes through 1e-10 @ 100 sec. and goes both up and down 1 decade for each time decade. Notice that your results turn down and to the right to follow this line. Different GPSDOs will take a slightly different path, but they all run parallel to the 'GPS line'.
Your Rb-Rb results are also about what's expected. The manufacturer only specs out to 100 sec. as 2e-12 and you're in that ballpark from 1000 sec. on.
Did you do the Rb-Rb tests first? I see that the first Rb-GPSDO test bottomed out at ~1e-12 while the second one kept on going down. Something improved between the first and second test. Did you make any changes in your setup, cables, power supply, anything?
The graph always shows the performance of the worst of the counter, the GPSDO, and the Rb. So below about 20 sec., your counter is blind to the performance of either the GPSDO or the Rb. All you're seeing is the counter's performance. At about 20 sec. the GPSDO's performance becomes the dominant factor. But the Rb is so good that you don't start to see it until you get to about 1000 sec. and even that may be misleading. The red line that keeps going down at the right side of the graph is following the GPS performance. The only way it could do this is if the Rb was better than the GPS. Otherwise, the red line would turn and run horizontal like the others.
So, if you really want to find out the performance of your Rb units, more investigation is required. You might discover that Rb 4 is somewhat better than the others. Such things do happen.
Ed