HFSS GCPW ports-only looks correct, full solution doesn't

Hello- Playing with what should be a simple GPCW model in HFSS. Wave port excitations, bilateral symmetry (perfect-H). Port heights are 4*dielectric; port widths are 3*(trace+2*gap). Ports-only solution at 4GHz gives Z0~50ohms and fields that look correct. Inspection of gamma for modes 1-3 indicates only mode 1 propagates. Full solution looks strange though. S11 is at -40dB @ 1GHz, increases up to about -10dB @ 4GHz. E-Field lines look correct near gap, but within about 2 trace widths E-field in the dielectric looks like a parallel-plate distribution between the bottom & side GNDs. If I shrink the airbox down by a few trace widths in x & y the solution becomes better behaved. Can anyone provide insight into what's going on and advise? Thanks!