The only numbers worth squat are the wind tunnel drag numbers. And those numbers, bike alone, are only 15% of the overall drag: when taking into account rider and bike and clothing.
In essence, - not much from not much leaves not much. Doubtless, that 15 pound Cervelo frame will do slightly better in wind tunnel tests than that 15 pound Colnago, - but given that, probable, 2% drag improvement due to the frame: there’s no real world rider test or tester that’ll be able to discern a difference.
Yes, but as the rider position and clothing are optimized, the % of the overall drag attributed to the frame will be higher. On top of that, for a “good” position, I think your 15% estimate is too low. For example, there’s a table in the Burke book “High-Tech Cycling” that shows the drag for a Hooker Elite w/disk and aero front wheel to be 1.38 lbs at 30 mph. With a rider on board (in an aero position), it’s 6.66 lbs. That means the frame is 20% of the aero drag in that case. According to the wind tunnel numbers revealed by Dr. Coggan recently, the P3C is as low, or lower drag than the Hooker.
BTW, the drag on the Hooker is up to 0.5 lbs lower drag than some of the other more “conventional” frames listed. 0.5 lbs of drag equates to ~2.5seconds per kilometer. I don’t know about you, but that’s pretty significant to me.
0.5 lbs out of ~7 lbs is ~7% reduction in drag. You may not “sense” it while you’re riding, but the stopwatch won’t lie.
Aero wheels will make a bigger difference than aero frames.
Nobody said any different…we’re talking frames here.
There was an interesting real world study conducted by Bicycle Quarterly, they found that “in the real world” low pressure, big old fat randonneur tires were faster than high pressure modern racing tires.
Yep…they’re faster when you’re rolling down a soapbox derby hill at only 13 mph 
You might want to take a close look at that BQ test again…there’s lots of “room for error” in their test technique. I won’t go into it here, but I wouldn’t take any more than just “broad brush” generalizations from their results.
Once you put a rider on a bike on a course, you lose numbers real fast and take our discussion, largely, outside measurably repeatable science. The difference being the way you attack a turn or two, or get out of aero to scratch your forehead…too many variables. Honest tests will reveal inconclusive results.
But, that’s not what we’re comparing. I said “for all things being equal”. That means the hypothetical case that you could replace one frame for the other and re-run exactly the same. The more aerodynamic frame will ALWAYS be faster.
Now, - those little wee wind tunnel numbers help the Cervelo marketing machine sell lots of bikes to slowtwitchers looking to get out from the back of the pack. My take is that they need to get out on their bikes…
Hmm…“wee little numbers”…heck, at least we’re making progress…it used to be “no difference”, right?
I don’t know, 2.5 seconds per kilometer comes out to be 1:40 over 40K…that doesn’t sound too “wee” to me.
I agree that if you’ve got a pinto engine, having a ferrari chassis isn’t going to instantly make you FAST. However, you’ll be faster than if you had a pinto chassis, though. And if you’ve got a ferrari engine (or even a well-built small block chevy) you’ll be even faster yet.
Here’s what it comes down to: For a given fitness level, speed in a TT or tri leg comes down to a BUNCH of little details. Obviously certain details give you more “bang for the buck”, but to state that frames make “no difference” is pure bull.
