How are you preparing your sample to go into the diffractometer? Try and get as close to a random orientation as you can. Use back- or better still side- loading, try using spray drying before loading, all of this will help reduce PO, then the software will have an easier time. Are you sure that it is pure montmorillonite and not interlayered? - if not do you have the correct structure?

Thanks for your answer bue the problem is that I am working with whole rock samples that aré not oriented

one approach might be to set the regions for the basal reflections for smectite to zero, i.e. don't include basal reflection in your refinement.  The smectite model in Siroquant is not necessarily derived form a truly random preparation which may be one issue and the other as mentioned by several people here is that your own attempts to rpeparae a random powder may not achieve it.  So although this will cut out regions where you may see obvious basal reflections by concentration the quantification on non-balsa peaks you will likely get a better result.  That said the Rietveld methods is tricky to try to apply to disordered clay minerals of whatever type and there are very few labs in the world that make a good job of it with any software.

Thanks Stephen!