Some parts of pretty much all the references still contain gaps, often due to having highly repetitive underlying sequence (e.g., around centromeres) or due to high heterogeneity between samples (i.e., multiple samples were used to generate the reads and they simply don't agree on a region, likely because the samples differ there). Assembling a genome turns out to be a rather difficult process.

Could it possibly be that you are looking at regions of unresolved sequence or assembly gaps? Regions around telomeres, and centromeres tend to be highly enriched in Ns.