MACS is best although you can try PICS as well.
Reason for using PICS: Two binding sites are separated into two disjoint regions by PICS, whereas MACS combines these two sites into a single region. This makes PICS particularly attractive for subsequent motif based analysis.
I have not had chance to look at this yet, but RSEG looks very promising! I expect this to be good as Andrew Smith has historically produced very useful software tools, e.g. CREAD and DME.
We have also had a good experience using SICER.
Although i routinely use MACS for ChIP-seq i would not recommend its use with histone data.
Maybe you can find the answer from this article: Shirley Pepke, Barbara Wold & Ali Mortazavi, Computation for ChIP-seq and RNA-seq studies. Nature Methods 6, S22 - S32 (2009).
And this question had been discussed in SEQanswers before, just try to google "ChIP-Seq Challenge".
I have not tried this myself, but you can have a look at the ChipSeqR Bioconductor package. There is a fairly extensive section on testing this method in the associated paper by Humburg et al. that may get to the issue of "best" peak caller for nucleosomes/histon modifications. This paper also cites two other papers by Albert et al and Kharchenko et al that are probably worth having a look at as well.