Common and interesting theme of discussion on many reloading forums. Having some spare time yesterday I dug back through years of my reloading data for multiple cartridges trying to get a sense for any crimp vs accuracy correlation. Generally, my data is both sketchy and skewed because so many cartridges, for proper function, absolutely require a crimp while some require only a specific kind of crimp with others requiring no crimp at all.
The only correlation between crimp and accuracy I noticed was within the “intermediate and smaller” caliber group of cartridges. For regular stacked magazine, jacketed bullet loads in bolt action rifles, I could not detect any difference in accuracy in either the 416 Rem Mag or the 450 Watts between crimped and uncrimped, all other things equal. I detected a small correlation in the 338-06, 35 Whelen AI and 375 HH with a very small loss of accuracy showing in the crimped group. This data tends to agree with that presented by the OP, but to a lesser degree. This correlation I would call marginally statistically significant. I could detect a slightly larger loss of accuracy crimping the 260 Rem and 270 Win and would say this correlation to be statistically significant. I never tried crimping bolt action ammo in calibers smaller than 260 Rem so have no data. All 223 ammo was crimped because of the requirement given its semi-auto use.
IMO, the correlation between crimping and accuracy suggests it may be caliber dependent. And, the priority for absolute function and reliability in the larger, DG calibers overrides all but the most negative affects on accuracy crimping may have.