It seems that you have the correct phase and the peaks lining up at right bond distances - so the overall fit seems quite reasonable!
Is the Qmax for your measurements 11.94 Å-1? That's a bit on the lower end for PDF analysis with sufficient resolution in real space. In my opinion, decent r-space resolution only starts building up after ~20 Å-1 . The lower Qmax might be one reason that you don't get a perfect fit and a weighted R of approx 35%.
As for choosing what parameters to refine and what constraints to place, this would have to be decided based on existing literature and physical reasonability of different models for your system. What is the question that you are trying to answer with the PDF analysis and what additional info do you wish to obtain as compared to the Rietveld refinements? Also, consider whether the effect that you are trying to investigate will actually show up in a PDF with the resolution of your measurements.
Hope this helps!