How to indicate expected binding orientation

I have been modeling several pairs of short dimerizing proteins. I know from previous papers on these proteins and experimental data that in reality they dimerize parallel. However, all of my top HADDOCK clusters show them antiparallel with the exception of less accurate models with RMSD in the 30s. Thus far I have been using the easy interface. Is there any way I can indicate the expected binding orientation in the settings so that I can eliminate models that dimerize antiparallel?

Remember docking is far from perfect…

Did you inspect the other clusters?
The RMSD reported in the result page only tells you were a cluster is with respect to the best generated model, but is no measure of accuracy.

Also if you dimers are symmetrical it might be worth applying C2 symmetry restraints and also NCS restraints