So if you clamp the fork and measure the handlebar deflection, but I clamp the handlebars and measure the fork deflection, we will get exactly the same answer (with opposite sign, obviously).
Are you certain about that? While the Newton's law is not under dispute, you can't simplistically apply it once to a complex system of points without being dreadfully wrong in most cases. In addition, carbon fiber is not a homogeneous material that responds to stresses in different places and directions in equal ways.
Thinking that newton's 3rd law will account for all of that seems problematic at best.
When undertaking finite element analysis, you have to make many assumptions about homogeneity of materials and response, as well as the geometry of each system. A simple example: if I apply a force to an isosceles triangle with 10cm legs, and apply a deformation to a similar triangle with 10m legs, there may be a slight difference in the deformation in the triangles right?
In addition, the force at the bars when riding is in a different direction than the force at the forks (i.e. your vector would be different) so if you drew out a force diagram even if we assumed that you could simply apply newton's law you will have to tell me how to best split up the opposite reactions as well, and that isn't account for by simply stating that if you clamp that it will show up there.
So by clamping the bar, you may find that the bb in fact swings more to the side, or that the seat tube/head tube junction does, or that the engineer who spent some time and attention thinking about how the rider might be pulling up on the bar, while pushing down on a pedal, while leaning the bike over a bit decided to place a few strategic plies here and there to combat one kind of torsion while purposely allowing other force vectors to deform the frame to ensure that the bike is comfortable.
Measuring that all at the same time just makes a muddle, as the testers found out: "Interestingly, the torsional test didn't correlate much with our overall opinion of a bike" (-p72 of the VN issue) I interpret this comment in context to mean that they knew the Ridley was stiff, as the tests point out, but stiffer didn't mean better to them.
I have a different take: while I agree that stiff does not equal better (it does not, and I love riding my old steel bikes, and I'm greatly doubtful that stiffness means more efficient either, until somebody can show me a study that demonstrates this, but that's a different discussion), doing a silly test of frame stiffness won't help understand the dynamics of the bike much, so why bother doing it at all?
Much better off using a slightly different design that mimics *one* scenario at a time, and perhaps being able to infer something from that.
I know these tests are never going to get as good as real life, but at least if we're going to be reductionist and test a certain factor, let's do that in a way that allows us all (and velonews most of all, if they want to make money) to actually be relevant.
IMHO it would be vastly superior to simply clamp the frame down sideways on a reference table, place weights in turn at each interesting place (BB, stem attachment, fork dropouts), and measure deflection than muddling around with pushing on one point and thinking that will tell us anything much worth knowing.
Somebody clever should actually think about the angles at which these forces are applied while riding and maybe tilt the table to account. And perhaps somebody more clever will make a bike riding dummy that can measure vibration and then we'll act (and spend) like the car industry