As far as I can tell (by running tests manually),
annotatePeaks.pl in the HOMER package  ignores strand information when searching the Transcription Start Site (TSS) nearest to each peak. After a close inspection of the various settings available, I couldn't identify one which changes this behaviour.
This was unexpected for two reasons:
- HOMER actually requires input to have a Strand column 
- The data from some NGS technologies are directional (e.g CAGE), so strand information should be used when assigning the peaks to transcript/gene models
Could someone please comment if I'm missing something here?