(1) Using a radiation boundary that is lambda/4 away from the transmission line should be fine. Since you are not simulating an antenna (strong radiator), you may could even use a PEC boundary if its far enough away from the transmission line. Since the signal is propagating down the microstrip line and not into free space, as long as you don't intefere with the fields above the trace it should be fine.
(2) Make sure the copper thickness and finite conductivity are correctly modeled. Also include the dielectric losses of the substrate in the model.
(3) Be sure to use a wave port instead of a lumped port, and make the port convergence criteria more strict.
(4) Renormalize the HFSS predicted S-parameters to 50 Ohms.
Good luck.