As a preface to my answer: we have added three new methods to the new release of MotifDb:
All three require a "source" argument, either "MotifDb" or "TFClass" - thus offering two data sources for the sometimes imprecise business of linking a motif to its cognate transcription factor/s. TFClass is a liberal source based on the 2015 paper https://www.ncbi.nlm.nih.gov/pubmed/25361979; multiple TFs are associated with many motifs. The MotifDb source is more conservative, typically associating just one TF with each motif.
For either the old (1.18) or new release of MotifDb, you can answer your question with this approach show below. Note that these counts use only metadata held directly in MotifDb itself, obtained from the providers, and is thus equivalent to the conservative source="MotifDb" approach described above.
length(unique(mcols(MotifDb)$geneSymbol)) #  3012 length(unique(mcols(query(MotifDb, "hsapiens"))$geneSymbol)) #  1273 jaspar.2016.human.motifs <- query(query(MotifDb, "hsapiens"), "jaspar2016") length(unique(mcols(jaspar.2016.human.motifs)$geneSymbol)) #  389