Ok, I've had a look into this. I don't own an Everdrive and therefore I cannot use the same controller test program. But I've got another device to read out the N64 controller. I noted the controller stick position values and put them into a Excel sheet. (Excel sheet attached to this post.)
1) At first I've tested my old GC style stick with a v2 Stick Converter PCB inside. It's the blue line and it looks perfectly ok.
2) After that I've assembled a v3 PCB set from the current batch. I installed it and the test results look pretty much like the ones shown by Seroczynski (red linde).
3) Finally I've desoldered the potentiometer stick from the v2 PCB and soldered it to the v3 PCB. The pins of the stick got a little bit bit bent during the desoldering process. After calibrating I've tested the stick again and now it looked pretty much than the old v2 PCB (green line).
My conclusions so far:
1) The v3 PCB's of this batch not faulty. They're working like they always have.
2) The accuracy/linearity of the potentiometer sticks seems to vary. But I don't know if the recent batch is worse than those before (or even out of specs).
I've been buying the same sticks
, same specs. Also the markings on the potentiometer sticks are the same (60, 6, B, 103).
Right now I'm guessing that it's normal that some potentiometer stick got a better or worse linearity than others. Seroczynski also said his other sticks got "better" test results.
3) There's no guarantee that after calibration you'll get a perfect octagon on your controller test program. Just look at the test results of other potentiometer-based N64 sticks, they can look really wild.
But even if the test result looks like the red one I've tested, then it doesn't necessarily mean the stick is unplayable.
4) @iacka: It's actually the first time that someone said he can't run at full speed in Mario64 with that N64 Stick Converter. Of course that's a real problem and should definitively not occur. I'll contact you soon via email to resolve that issue.
5) I got an idea for different calibration method. Maybe that *could* tackle the problems that occur when the potentiometer stick's linearity is worse than it should be. Unfortunately I don't have time for that at moment, but it will give it a try for sure.