The common wisdom is that CHARMM forcefield is "better" for proteins while AMBER forcefiled is better for DNA simulations. However I think the answer is more complicated than the aforementioned statement. Look at the following, it might help you to understand some of the differences:
1- Assessing the accuracy of physical models used in protein-folding simulations: quantitative evidence from long molecular dynamics simulations (http://www.sciencedirect.com/science/article/pii/S0959440X13002157)
2- Comparison of Secondary Structure Formation Using 10 Different Force Fields in Microsecond Molecular Dynamics Simulations (http://pubs.acs.org/doi/abs/10.1021/ct300323g)
3- Are Protein Force Fields Getting Better? A Systematic Benchmark on 524 Diverse NMR Measurements (http://pubs.acs.org/doi/abs/10.1021/ct2007814)
It is doubtful that the CHARMM is "better" for proteins. At least my attempts of the protein folding in CHARMM have not been successful, while in AMBER were successful examples.
That is why I said "better" not better and the first reference I mentioned reflect your point about folding. But I am not sure that from limited number of examples one can make a final judgment.