Passive S-parameters from Active for Two-Element Array?

I’m trying to model a two-element dipole array that can be steered in azimuth arbitrarily, and I want to compute realized gain.  To do that, I run two FEKO simulations, one with in-phase excitations, [V1 V2] = [1 1], and one with 180-degree out-of-phase excitations, [V1 V2] = [1 -1].  From FEKO I export far fields and S-parameters, then process in Matlab.  I’m assuming that the S-parameters that Feko calculates are active S-parameters since Feko computes only one value for each port.  Is this a correct assumption?  Assuming it is, I calculate the passive S-parameters from the four active S parameters (S1 and S2 for both [1 1] and [1 -1] excitations).  I then use these to calculate active S-parameters for arbitrary element phase, and use the calculated active S-parameters to calculate the reflection loss needed to compute the realized gain for the array.  Unfortunately, my calculated results don’t agree with those from Feko.  That is, if I steer the beam 45 degrees in Feko and in my Matlab code, the active S-parameters don’t agree.  Have I made an incorrect assumption about how Feko operates or is there a flaw in my logic?