Interesting results.
So I disassembled the base and loosened stopper 3, used two hex keys to rotate the metal disk clockwise as you requested and then re-tightened the stopper.
Here is what it looked like after:
Sorry it is sideways
I reassembled the base, reattached it to my PC, set it to default and re-calibrated. After some messing around, the issue began to show up again. So I stopped my testing when it started repeating the previous pattern (hold left a little and forward a little for a while and then move forward and backward with the bumps showing up in the axis plot) and carefully disassembled the base again.
Note the two marks have gotten closer:
I tested the stopper and although it was not loose, it was not as tight as I had set if minutes before.
So I re-tightened the stopper, attached my KG12 grip that came with the base and without reassembling the base, moved the mechanism around both axis a fair bit (a minute of tinkering) and took another look at the marks.
The marks are almost back to where they were and the stopper is not as tight as I had set it a minute before. Granted, I am somewhat cautious about over-tightening the stopper.
I repeated the test again (with the KG12 and the base not reassembled) but this time I tightened the stopper and then manually loosened it to ensure that I was tightening it 'enough' and it was requiring some torque to loosen it by hand. I then tightened it again, added the KG12 stick and moved the mechanism again for about a minute. When I checked the stopper it was, again, just barely tightened and it required much less force to loosen it than when I did tightened/loosened it just a minute ago.
It seems like the stopper is not holding very well and with a few minutes of motion, it becomes virtually loose, requiring minimal force to loosen.
One other thing I noticed when calibrating the stick was that the Y Axis (front to back) , it was difficult to get it to show 0 - 65535. It would do it in the calibration but by the time I ended the calibration and testing the axis setting while observing the values in the VKBDevCfg-C.exe app it would show only 1xxx - 65535 or 0 - 64xxx ... like it was unable to register the full motion. After a few attempts to claibrate, I finally had it showing 0-65535 before I began my axis plot test with the VKB_JoyTester. At the time, I do not think that the disk was loosened again and back to it's position where the two marks line up but I now figure that the stopper was probably back to it's pretty-darn-close-to-loose state. Then, when I was doing the testing with the axis plot, the disk started to slowly move. Not sure if that helps.