Well, as I said, the angles in the drawing are exaggerated for the sake of clarity - I guess my beef with your argument is that I can't see any particular reason why lift must act through the CofG.
To put it another way, what if we had a helicopter hovering with skids level, for the sake of argument.
If we had a rope tied to the right skid and a bunch of very strong guys pulling the rope directly outwards, they would move the helicopter in their direction.
To stop the movement, the pilot would put in enough left cyclic to oppose that pull exactly.
Now (I think, anyway) we'd have a situation a lot like in my diagram - a force to the right from low down (rope on skids) and one to the left from up high (component of rotor thrust going left), stopping drift but creating roll.
You wind up with the CofG displaced to the right from under the head slightly. The harder those guys pull on the rope, the more the pilot opposes it with cyclic (having to also increase collective because his rotor thrust vector is more tilted too), and the more roll you get, with the CofG moving further out all the time.
The component of rotor thrust going straight up still equals weight, holding the helicopter in the air, but the opposing lateral forces from high and low want to roll it.
It stops rolling when the CofG gets far enough out to generate an equal and opposite rolling couple.
I reckon it's a bit like a punching bag hanging off a hook in the roof - say you pull it sideways at the bottom with a small but constant force. It will move towards you until the opposing force caused by the CofG being displaced away from under the hook is equal to your pull force, and will then stay hanging there stationary with the CofG out to one side because you're holding it there.
May have to call Professor Julius Sumner Miller in for an adjudication.