A question often asked is “which options work best” to which the answer is always “it depends”. This is primarily because the outcome of the analysis depends more on the quality of the data than anything else. However I would always try for yourself and get a feel for how the program works for your data - running both -2d and -3d will simply require more computing time / disk space rather than more effort, so it is certainly worthwhile. For small molecule data though -3dii -small molecule is a good mix. Also -3d often works better for very finely sliced data.
Users of xia2 are asked to cite Winter, G. Journal of Applied Crystallography (2010) 43, 186-190. Please also remember that the programs that xia2 has used must be correctly cited, which are reported in the program output.