<CST> Defining Ports and B.C. for a power splitter &lt

Hi All,
Background story:
I'm designing a 2-way power splitter and would like your comments/advices about some issues i've encountered.
First of all, the splitter is a "standard" wilkinson power splitter, using a 50 ohm microstrip line on a grounded dielectric slab.
After building th model, adjusting the mesh and defining waveguide ports, i've conducted several simulations, each with different Boundary Conditions (B.C.).
1st. All electric BC ( as CST recommends for this type of models )
2nd. All open BC (Tight)
3rd All open expanded,
4th All electric with full de-embedding.
The problem :
Each simulation gave different results from the others !
I understand that there's a difference between theory and "real world" and some optimization has to b made, but....the same splitter resonates at 4GHz in the 1st and at 5GHz at the 2nd simulation....
Can you help ?
BTW: I made sure same mesh density and port impedance is kept , especially at the 3rd simulation.

Obviously BCs do matter and obviously changing the BCs affect the resonant frequency. In real world the splitter would go into a package and hence CST recommends using Perfect E BCs. I do not have any problem with your results.
-svarun