I would also suggest using the AMOEBA force field. Strongly polar interactions can be reproduced with improved accuracy in the electrostatic term relative to simple point charge models (AMOEBA includes a charge, a dipole and a quadrupole on each atom). Polarization can be expected to be important as well (e.g. for heavy elements in the cases you suggest) and in such cases the self consistent treatment in AMOEBA will be useful. One term which will be lacking in most FF, including AMOEBA, is charge transfer.
The answer depends on the physical state in which you want to do your calculations. In water solution you get salt dissociation. In solid sate you do not have to face this issue. Nevertheless, AMOEBA could be a good bet if computer time is no problem
Forece Field proposals are fine. I recomend you to test different simulation parameters with the known structure most similar to your system, in order to find the best conditions for energy estimation.
I used OPLS-AA on Schrodinger MacroModel with good correlations with experimental data on a- and b-cyclodextrin complexes with n-alkyl carboxylic acids and n-alkyl p-hydroxy benzoates. A molecular mechanics study of 1:1 and 1:2 associations, Antonio C.S. Lino , Yuji Takahata and Carlos Jaime. It is very interesting to see the hydrophobic effect with the along of carbon numbers.