are you making this up, or is there a study you can link to which supports any of this?
Wider tyres having a lower Crr for the same outer diameter has been known for years. Wide rims make the effective width of the tyre larger & thus reduce the Crr.
With my training bike in Italy careful selection of data allows me to accurately estimate the combined front/rear thickness of my rim sidewalls from Crr data! When building wheels in forks rather than a a truing stand I can see how egg shaped the wheel was &/or I can see if the spoke tension was low/high.
I can see the difference between my old (1.9mm spoke seat) & new (3.6mm spoke seat) chinese carbon rims, the newer ones which are noticeably less prone to radial distortion give a lower Crr than the older ones, especially towards he end of their life when cracks started to develop around the spoke holes.
On my fixed bike the Crr data can tell me if I need to true the wheels up, the rims take a lot of punishment as it's harder to change lines quickly to avoid small potholes etc. I can also tell if I was riding with the old or new rim set through the Crr data, I couldn't tell when I laced the new rims onto the older hubs (which are slotted for aero spokes) & the old rims onto the new hubs from Crr data because it was identical.
You can do your own research if you want to. You need 4 things:
1) Accurate speed measurement with a data recorder that gives high resolution sampling (1s is fine, but 0.5s is better). A Garmin Edge 500 does a fine job of this
2) A descent with a known accurate distance & drop, the profile of the descent doesn't matter as long as you can freewheel or soft pedal all the way down it (This means your power output is know to be 0w)
3) A still day or your descent to be sheltered from all wind.
4) A spreadsheet to input your data into (see this
post on another forum for a working spreadsheet)
You can go further than this by using a calibrated power meter & roads with accurately known ascents/descents. You can work out your aero & rolling drag to confirm roll out data. Something like aerolab in GoldenCheetah can become invaluable for work like this. The key to all of this is choosing your data sets carefully.