Check out Frameless Backpacks: Engineering Analysis of the Load Carrying Performance of Selected Lightweight Packs and Quantitative Analysis of Backpack Suspension Performance by Dr. J. Aside from hopefully helping to answer your question, they are fine examples of BPL's roots.
The first article clearly shows that a McHale Subpop with stays clearly out performs one without in terms of torso collapse (the indicator used to measure load transfer) at total weights above 20 lbs, at least when the packs have a "soft good" density of 2.1 oz/L supplemented by weighted metal discs.
Ryan's analysis, however, focuses on claims that a rolled cylinder adds support, and pack density was not a tested variable. At first glance, the density (which would work out to about 6.6 lbs for a 50 L pack) doesn't sound like much, but you have to remember that some goods (like water, food, cooking gear, tent poles) aren't soft goods, so aren't included in that number. Unfortunately, the metal disks used may supplant the weight, but not volume of those other goods, so actual soft good density in a real world 50 L pack with 6.6lbs of soft goods may be somewhat higher.
The problem that I see with trying to increase soft good pack density to increase weight transfer (which may or may not work), is that at some point you may be damaging the lofting abilities of those goods.
If this question is really burning for you, I would by all means recommend taking his methodology, and adapting it to measure the effect of soft good density on torso collapse. I'd actually be quite interested in seeing that!