The case length is related to how much crimp is applied. That's why screwing the die in or out changes the amount of crimp, too... a crimp die converts vertical intrusion into the die into a horizontal crimping force. That's what a crimp die does. So if you have big variances in case length, you can get big variances in the amount of crimp.
Ordinarily, this doesn't have a lot of practical effect on most taper-crimped pistol cartridges, which is why few bother trimming their pistol brass. I think something else is going on.
Here's my hypothesis: It's the chamfering. You are thinning the brass so much that, in combination with a somewhat-long case, the horizontal crimping action is sufficient to actually or nearly shear through the brass. You are either cutting off the tallest bit of the brass (because you have thinned it by chamfering) during crimping, or you are setting it up to shear during firing. If it's not actually being fully sheared on the press, the combination of chamfering and crimping has gotten some brass to a degree of thin-ness that, under pressure in the chamber, the last little bit separates.
You don't need to chamfer pistol brass (especially "inside and outside" - why?). Pistol brass is crimped, and semi-autos, including 9mm, generally headspace on the rim. The rim is doing work at full thickness, and is also being worked in a way that rifle brass isn't. Don't thin that out... it's structural.