Its hard to say without looking at the result in detail. Is there any sampling of closed conformations at all? Does it converge into a particular dimer structure? This behavior would be expected if the interactions between dimers is not sufficiently large to converge to a particular dimer-dimer orientation.
The x(1) z(1) does not make much sense because the subunits would end up on top of each other at the beginning of the simulation. You could play around with alternative ways of sliding subunits together. In the make_symmdef_denovo.py script you can control the slide behavior ( -slide_type (RANDOM|SEQUENTIAL|ORDERED_SEQUENTIAL).
If there is a good convergence for a dimer you can try symmetric docking of these dimers followed by a symmetric relax.
You could play around wit