Note also that in "factor" mode, HOMER produces fixed-width peaks.  For very strong peaks, that may mean that most of the reads are outside the range of the called peak (e.g. it gives you 200bp wide peaks, but the bulk of reads in a very strong peak might be more than 100bp from the centre).  If you're looking at differential binding, this can give misleading results, since reads that are off the edge of the called peaks won't be counted.