What Kristopher said is essentially the same thing I heard from a Solyndra official one year ago. I was at a conference on sustainable energy and when they described the Solyndra design my first thought was that it was fairly inefficient for the obvious reasons. In terms of electrical output per unit solar cell surface area in full sunlight, you simply cannot beat a tracking system.
But that's probably not the most relevant metric; a better one may be installed cost per kWh generated. Tracking systems add cost and complexity. The not-totally-crazy idea behind the Solyndra design was to make some tradeoffs. You get a lot of mechanical simplicity from a non-tracking system, it does vastly reduce potential limitations due to wind and (in cold climates) snow load, and allows collection of some diffusely-scattered light from surfaces beneath the tubes (i.e. the half facing "the wrong direction" . The design simply isn't intended to maximize output from a given roof surface area; instead, the idea is to get some power more cheaply.
But the whole concept really only works if you can put a sufficiently cheap PV material on the cylinders. For a while it looked like they might have an edge there, but then the economics of solar panels changed abruptly when the Chinese panels flooded the market. It probably was a pretty marginal concept even without that; but cheap silicon cells meant no chance.
I'd love to know exactly when, where and why DOE gave the system a big thumbs down, aside from any shady dealings and market shifts. Just providing a link would be great.