Sounds to me like the automated calibration procedure is kinda meh, and a dodgy cell is compounding the problem. Assuming you're actually using 100% O2 and not some concentrator garbage or the like.
Cell "wake up" is a thing when you first take it out of the package, because they're stored in sealed bags full of inert gas. I leave mine for an hour then put it in my mini cell checker with O2 to get a baseline reading before installing them. That's really the only time that it should exhibit that sort of behavior. If it's exhibiting this type of behavior outside of the package, you've definitely got a dodgy cell.
Run all 3 in the cell checker to see what's going on without any of the electronics involved. It certainly sounds like the calibration procedure just assumes that you'll get a good purge, so when you have some lower FO2 gas "contaminating" your purged loop, it jacks your results up. Like has been pointed out, your numbers are out of scope. If your numbers are good in the cell checker, but they're not in the unit, then you've got a unit problem. If they're bad in the cell checker, then they're not gonna get better in the unit no matter how solid the unit is.
Personally I don't like automated calibration routines because they make assumptions that may or may not be true. Garbage in, garbage out type of thing. My Pelagian is literally two pots in the handset that you use to adjust the displayed PO2, just like the Analox units. It's about as dead simple and accurate as you can get. The SF2 has a cap that screws on one of the hose ports and blows O2 straight across the cells. You don't hit confirm until the values stabilize, and you can see the mV vales during calibration, so you can make sure your numbers are where they're supposed to be. It doesn't do anything with the solenoid, doesn't require complete purging, etc. It's got an LP port on the top of the cap so it provides positive pressure, making "contamination" a non-issue. I dunno if the JJ will let you do this, but it's worth looking at.
If you can find a decent way of calibrating with just the head disconnected from the rest of the body you'll use less gas, and you'll be able to control the variables a little bit easier. If you can find a Tupperware bowl that's the right diameter, they make great "head only" bottom halves. Couple that with some sort of cap with an inlet that lets you blow O2 straight to the cells and bypasses any automatic solution, you're more likely to get as accurate a calibration as reasonably possible.
Either way, you know you need a new cell, but it's worth doing a bunch of fiddling to see how it's behaving in difference scenarios.
