Ummmm,
I had the same problem the last year when i was designing a formula student car steering system... I was new doing FEM analysis and if i see the results, maybe i would be ashamed!!! hahahaha.
Now i am more experienced and i would approach the problem as follows:
Put the boundary conditions in the unmovable side, make a set with the nodes and select U1,U2,U3.
To simulate the torsion, instead of creating a momentum in the face, create a distributing coupling to put the momentum in a node.
Master node: A point created in the middle of the cylinder face.
Slave nodes: The nodes of the face.
When you select the distributing coupling, could be enough to active only the spin that affects to the momentum. Do some tests.
Now that i have time i will try your model. Give me some dimensions and a load and i will try.
Regards.