It doesn’t matter.
On a publication I know, and according to a specified Reynolds number. Other publications that I have cited give other values.
Please read all the links and comments I provide on this issue before making an opinion.
The 4/27 limit for yo-yo mode is logic: the swept area goes downwind, the apparent wind speed being 2/3 real wind speed. The paper proves it. My demonstration ( I put again below) proves it, using the formula of force multiplied by reel-out speed to match the formula of power at Betz limit. This has nothing to do with the drag coefficient. This is related to the loss of apparent wind speed, given that the swept area is moving downwind.
That said when the Cd increases beyond 1, we can reasonably assume that the area swept increases also (see my previous comment), being the parachute (or autogyro) area plus the area of air deflected by the parachute and which increases if the Cd increases.
Thus, the demonstration remains valid even if I use a Cd of 2 (the disk area becoming 200 m² by taking account of additional air deflection), which I do below, with a temporary conclusion.
Note: 4/9 is (2/3)². It is because the traction force goes with squared wind speed.
Now, with a Cd of 2: the traction force is 12000 N instead of 6000 N. The result is: 40000 x 4/9 = 17776 W, so exactly (to within decimal places) 1/4 of 71104 W, by taking account of the doubled swept area, which becomes 200 m² due to the additional deflection of air around the parachute or autogyro.
In practice this would mean that the higher the Cd is beyond 1 or even 2 (a value much higher than that from tests on a parasail that was not a scale model linked here), the more the units should be spaced out to maintain their aerodynamic efficiency.