Now that is interesting. I have always believed that chainfires can happen at either end, by penetration of the flame. and I still believe they can do so some of the time. Long ago I used to shoot an original London Navy, which on the basis of serial number and the government's broad arrow mark had some chance of having been down the valley and back in the Charge of the Light Brigade. I tried to work it with the Lee conical bullet and just the lube in the grooves, and got chainfires solely in one chamber which had become oversized. It wasn't a really loose bullet either, for it didn't creep forward under recoil.
I always mistrust experiments which may not have been repeated often enough for total validity, or with other variables eliminated. The reason for something not happening demands better statistical validity than the reason for something happening. But the non-happening of chainfires with a loaded and uncapped cylinder seems pretty compelling. I don't see any reason why 300 degrees of the breech face couldn't be slightly relieved to prevent contact with the caps. But it might be that gas impact on the caps was doing it.
There used to be a dodge employed by waterfowl shooters, of having a piece of soft rubber tubing over cap and nipple for waterproofing. (I would think of it as bicycle valve tubing, which dates me.) If chainfires at the rear really are due to penetration of the flame, this would surely eliminate it. At the front, soft grease which goes down into the crack between ogive and chamber wall surely would.