-
Lenticular is probably the best shape a disc could be if it were a separate body
Depends on the yaw angle. Flat is likely to be better at ~0° simply because of the smaller projected area. Once you add a bit of crosswind, the lenticular ones can start acting as sails and generate enough lift to make the resultant force on the bicycle act forwards; it's not strictly a free lunch as the other component is a large sideways push which has to be resisted by some slip angle at the tyre, which adds rolling resistance, but that's likely to be worse with a flat disc too, as it will tend to stall much sooner.
-
Pro Tour teams like to plump for Lightweight discs - a disc that is not lenticular but conical (due to how it's made)- so presumably it is 'fast' because it's stiffness and really light weight trump its shape.
I wouldn't rely on a lot of ProTour teams or riders to have much science to back that choice. 90% of riders in a TT stage treat it as a semi-rest day anyway.
Pro Tour teams like to plump for Lightweight discs - a disc that is not lenticular but conical (due to how it's made)- so presumably it is 'fast' because it's stiffness and really light weight trump its shape.
Lenticular is probably the best shape a disc could be if it were a separate body, but like @umop3pisdn says, if a frame shields the wheel well enough, it's shape matters less, and most modern frames shield them pretty well (if not as well as Cervelo).
The way it interacts with the frame is probably where the gains/losses are (and that's also why Zipp discs have [mostly] always been flat - see last paragraph:
http://www.zipp.com/technologies/aerodynamics/boundarylayer.php ).