There's lots of debate on which frame is most aerodynamic; the debate itself is not helped by the publication of manufacturers' tunnel data, which often relies on a series of shaped assumptions and protocols. What's needed is an agent who would test the frames independent of manufacturer involvement. This is a very difficult proposition, even apart from cost--there still is the issue of how variables between frames can be normalized so that meaningful data can be collected (for instance, how should the tunnel test for different height head tubes among frames that are nominally the same in size?). An individual on one of the large discussion forums has decided to independently test a wide range of time trial/triathlon bikes, with the support of the forum community members, who are loaning bikes for the cause.
It has been my practice to isolate major variables first, and then look a finer details of granulation. Here's what I've suggested:
I know that little differences when taken cumulatively can yield significant performance benefits, but another way to consider testing is to start with a broader cross-section of design categories, get some trend data, and then test specific makes under the best performing category. This should save you some time (no pun intended) and money in your initial testing scenarios.
From a UCI legality point of view, there are really two primary design variants to consider: frames like Look's 596 (large wheel cutout) and frames like the P4, Shiv, etc. I'd also throw into the mix Kestrel's Airfoil Pro--the 2009 version which is greatly refined over prior versions--for another set of data points, as well as a beam bike and a standard 1 1/8" round tube steel frame . Essentially, what I'm suggesting is that you try to determine an optimal design category first and then move to a final level of granularity in terms of testing a broader range of bikes.
Also, I'm starting to suspect that small changes in the front end of the bike makes huge differences; I'm wondering if it is possible to isolate the impact that the fork has on overall performance. If not, then will the standard be "as manufacturer spec'd," which can introduce a whole other set of variables. For instance, if frame A tests marginally slower than frame B, is the difference attributable to the fork and would a swap in fork make an appreciable improvement? Cervelo apparently did this when they replaced their Wolf fork; I know that adding a Blackwell fork to my P3C made a fast bike even faster.
Bicycle aerodynamics is an incredibly complex phenomena that is dependent on oftentimes individual variables. It is important to remember that the benefits of even the most aero frames can be negated by poor body position or equipment choice. It will be interesting to see the results.
Australian Superbike: http://www.powerhousemuseum.com/hsc/bike/article.htm