This question got a lot of views but no answer. Below is the solution I found. One can simply use any TSS file (or any reference file for that matter) as a HOMER peaks file or .bed file, as the peaks file against which your reads/peaks of interest will be mapped, and just ignore the tss function of HOMER entirely.

$annotatePeaks <custom_tss.bed> <genome> -d <tag directory containing reads> -options > output.txt

in this example the customer tss file was a .bed and the ChIP-seq reads were in a tag directory made by HOMER using makeTagDirectory and the genome is the genome as downloaded via HOMER.