Not sure about wall thickness, I know that the finish inside dimension is
0.623 +.0015 and -0.000.
I don't think the tube would deform as the clamping pressure of a 5C collet is not enough and 5C collet won't allow dimension over 0.750, at least not the ones I tried.
Jeff
An Ortleib Quadra that can use current-production rubberflex as well as steel pad and steel sub spring-collet would be serious spendy, but if it fit your spindle, might make lots of work easier.
Power operated, of course, so it plays well with CNC cycle times.
Among the several advantages it could bring is ability to hold a matched pair of rubberflex, back-to-back, reversed taper-cone direction.
That gives you extra grip against push-back - rather a LOT of it, compared to even a perfectly fitted 5C - and whether the tubing deforms to better roundness or not. or remains at a no-longer critical matching 'clock' position.
I have 'manual' Rubberflex but would not tackle 3,000 piece runs with it unless I also got to magically be 20 years old again as part of the deal.
Good TIR, forgiving fit, good grip, but RPITA to open and close.
BTW .. you've cited:
.. maybe 0.746 for the smallest dimension and 0.753 for the bigger dimension.
As it HAS that variance, I'd expect variance within variance, so with any 'spring' collet, you'll have to align the stock by hand and/or metrology, hope it stays within range of your custom 5C, self-aligns as the feeder cycles.
Trust that to a three-leaf 5C? I do not.
You can appreciate why I'd assign my Rubberflex or Burnerd Multisize to this sort of tasking, as neither of them gives a damn about that mere nuisance-level variance.
Shorter life of more 5C segments and less concern as to how the stock feeds might be more sustainable, overall.
Turn a 3-lobe into a 6-lobe, suck it and see. You'll have an answer faster than you can get a custom grind or EDM done.
JM2CW