Significant discrepancy between ADS and Momentum simulations

I designed a simple 90 degree branchline coupler in ADS, tuned it to my liking (my values were very close to theoretical). Now, I simulate the structure in Momentum and my curves are all off, I get 3 dB less since it got shifted by several GHz.
Any suggestion and ideas why might my results be so different (and which one to trust)?

http://imgur.com/a/1cIr8

http://imgur.com/a/1BDYJ

http://imgur.com/a/pMq19

Schematic and Momentum simulation result may be different. If return loss is 3 db less then it's quite possible but insertion loss should be reduced such an extend .

Correction
insertion loss should not be reduced such an extent .