What you are not accounting for is the fact that all 3 plots are taken at
3
different heights in FEET. So the plots showing a 3.5 MHz loop at say 70'
is
about 1/4 wavelength high, but that same loop at 14 mHz is a full
wavelength up-
this accounts for the lower takeoff angle, not the increased length of the
loop.

Good point, height has to be taken into account, except at 160 meters, since
a full wavelength loop has a 90 deg. take-off angle (cloudwarmer)
independent of height. A 2 wavelength loop will have a lower take-off angle
regardless of height (the higher, the lower angle).

At medium frequencies, it is the height relative to wavelength which is the
dominant factor as you point out. Scaling of the examples in
http://www.cebik.com/atl1.html shows this also:
- 80m loop @ 7 MHz and height 75' = 22.9 m = 0.57 wavelengths: elevation
peak at 26 degrees
- 80m loop @ 14 MHz and height 35' = 10.7 m = 0.54 wavelengths: elevation
peak at 26 degrees
Same take-off angle, at approximately the same relative height.

At high frequencies, the following statements from Cebik concerning the 80 m
loop relative the 160 m loop can still be extrapolated to 160m vs. 320 m
loop, I would say: "One might well argue for some installations that the
benefits derived on 80 meters from the larger loop are offset by the
disadvantages on some of the higher bands." "There is a strong possibility
that, if your interests are in upper HF operations, the large 160-meter loop
will prove to be a disappointment. Its true virtue lies in the lower HF
region, especially on 80 meters, with reasonable good performance through 20
meters." "Although the 80-meter loop shows poor performance on 80 meters for
every application other than NVIS, the smaller loop has distinct advantages
over the larger loop on almost every other band."

Sverre, LA3ZA