In Reply To:
I've done both many many many many times on various different bikes. Bottles still matter except on really bad DT bikes. I've tested pretty much every combo of bottle location/size on aero bikes over and over with and without rider.
As with every rider test, it's tough to get good repeatability (usually +/- 15 gF if done back to back) but you can still measure a difference with a rider, with and without bottle. So yes, DT round bottles are measurably worse than a bare aero bike (with a good DT).
Aero bottle results vary per frame. We designed ours to work well with Transition and the key competitor frames (P3C, TTX, DA, etc.) but not to the level of integration that Cervelo has done with the P4 bottle. We effectively wanted a bottle that head on would not hinder the bike but in crosswind could potentially give some lift. That's what we've ended up with -- a good usable bottle that meshes well with all kinds of frames.
Here's some data since we're on ST:
Preface: Shiv bike alone, Roval Rapide wheels w tubular tires (spinning), head-on drag normalized to 30 mph, rho=1.20; my typical protocol
Note, not yaw data here.
Bike, no bottles: 537 gF
(No pic, but same as below without bottle)
Bike, DT 21oz center of tube: 556 gF
Bike, DT Aerobottle, 1 cm gap to ST: 534 gF
Whats the time of savings of this 20gF advantage on a 40km TT?