weird HFSS simulation results for a simple wilkinson divide

weird HFSS simulation results for a simple wilkinson divider. it's known that the branch line's characteristic impedance is root 2 times the main line's characteristic impedance， so for microstrip line, the main line's width is bigger than the branch ones, but in my HFSS modle, the insertloss and isolation get food performance when the main line's width is smaller than the branch ones ,which means the main line's characteristic impedance is bigger. that's oppersite for the theory. can anybody figure out my errors in the HFSS modle below. HFSS 11 version
Thank you with my best blessing

any one snd me hfss simulated wilkinsn power divider pls