(my real name is Todd)
First, I'm not saying that a carabiner plus knot is always better...I'm saying its easier than a knot alone, and possibly less error prone. Mostly.
Also, in TESTING the fewer variables the better, If you want to test the fall factors of a rope alone....why use a carabiner and potentially invalidate the whole test?
IMHO A knot plus carabiner IS simpler than a knot alone. I base this on 2 very un scientific observations. 1) People are lazy and default to whats easiest and most cavers default to biners. 2) I generally find tying a knot around something harder than tying a knot then clipping. With sufficient practice and a larger "library" of knots I was comfortable with then I'm sure I (and others) would get better. But in the here and now biners are easier for me and almost certainly safer for my patient.

I'm also basing this on the "pool" of likely people involved. So yes, if I had 30 NCRC knot gurus on a rescue (or Hamilton County

If we are just tying on to a mobile patient in a harness and hauling them straight up and out...sure the simplicity difference between the two drops a lot. If we have multiple pitches seperated by horizontal passages in a litter...knots alone could suck.
Climbers may understand *climbing* falls better. I'm not convinced that your average climber fall vs average (if such a thing exists) caver fall, or even belay are analogous. You seem to imply that because of the use of static rope in caving its more important to prevent equipment failure...I'd say equipment failure would be a bad thing either way

As for a tumbling fall on a carabiner of unknown orientation...yea that could be bad....so could a screwed up in-line knot....you makes your choices and takes your chances.