Ok, you are starting to convince me 
In the meantime I tried a laminar flow assumption on the NON layered mesh with air viscosity 0 (I was trying to stop rotation) but that only got the Cd by CFD to be only 320% bigger than XFLR5.
I will try the 0-time look at NON layered mesh with potential flow initializing and report back.
2D is next
BUT for sure my eyes are opened and are seeing for sure why XFLR5 type programs should not be relied on for real life 3D situations about dragâŚ