An extra layer of information that is now available and worth considering is DNAse I hypersensitivity and digital footprinting data (which has been generated systematically in a variety of both human and mouse cell types by ENCODE). The idea is that in a particular cell type, regions that are transcriptionally active will be accessible and sensitive to DNAse, and at a fine scale, the actual sites bound by TFs will be protected.

See these papers:

• An expansive human regulatory lexicon encoded in transcription factor footprints Neph et al. Nature. 2012 Sep 6;489(7414):83-90 http://www.ncbi.nlm.nih.gov/pubmed/22955618
• Circuitry and dynamics of human transcription factor regulatory networks Neph S, Stergachis AB, Reynolds A, Sandstrom R, Borenstein E, Stamatoyannopoulos JA. Cell. 2012 Sep 14;150(6):1274-86. http://www.ncbi.nlm.nih.gov/pubmed/22959076
• Predicting tissue specific transcription factor binding sites Zhong S, He X, Bar-Joseph Z. BMC Genomics. 2013 Nov 15;14 http://www.ncbi.nlm.nih.gov/pubmed/24238150

Unfortunately there's no actual answer to your question. Whether a given binding site is actually used will depend heavily on cell-type, developmental context, and the nearby chromatin state. It could be that the genes with 100 binding sites in their promoter regions are never modulated by the transcription factor you're interested in in the context that you're looking. The only way to know would be the normal wet-lab experiments (and some promoter-bashing if you want to really know which site(s) is/are being used).

This question is impossible to answer with bioinformatics alone. Even ChIP-seq can potentially mislead you. Depending on the trancription factor, you could potentially have a binding event regulating a gene that is far from the nearest promoter. There is often a correlation between TF occupancy and nearby gene expression, but there are certainly long range interactions of importance in higher eukaroytes.

You could check to see if anyone has used RNAi to knockdown your factor in the animal or any cell line while monitoring occupancy using ChIP-seq and expression. This combination of data could be more informative in deciphering which sites may be occupied in a given context, and what genes the bound factor may be regulating.

The genome wide DNase data could be telling, particularly if hypersensitivity flanks a large number of your motifs on genes your hypothesize may be regulated by your factor, especially in a way that correlates with expression.

Bede