I did a bit more research and i figure out that for forces which act normal to cut planes, like torque, cyclic symmetry should be used. With cyclic symmetry used instead of regular symmetry i obtained result which is more / less in line with "full" pipe analysis - 2,830 psi. Difference is about 23%, which is still substantial but bearable as the plot itself makes sense with expected stress distribution.
i would still like to see if i am making any kind of mistakes in defining the model that i cannot see at this time.
When the mesh changes the results can also change. So refine the mesh untill you will get a constant stress value. It should be the same value in the symmetric and nonsymmetric case. In your case the size of the mesh is not identical, and so the results will be also different.
Actually, I believe, one should refine the mesh until the strain values level off. Stress may not be a good indicator.
If one used the same mesh size and type wouldn't you expect the same or similar results? The symmetry should mostly serve to reduce the computation time and not (greatly) affect the accuracy I would think. Is that not so?
I figured out one part of that problem.
As we know stress = F/A, so if i reduce the area where my force acts, i should reduce the force as well. That should take care of this non-issue. I believe that torque should stay untouched