Update: I was able to follow the code from QUCS to implement the complex impedance transmission line via Y parameters in the frequency domain device and it seems to work and give the expected loss. Schematic attached. I guess the RLCG model in QUCS is just doing this. The loss should be wiggly vs length because there are standing waves on the line and certain regions have more I^2R/unit length than others. The normal Tline does not know about this apparently and it seems to over or under estimates the loss.