I guess the easiest would be to compare (e.g. in UCSC genome browser) whether the region masked (e.g. by blatting the non-masked part) corresponds to an element in the repeatmasker track. (Unless you need to check more than a few parts and would like to automate this.)

Probably most N's in scaffolded assemblies are due to gaps between assembled contigs where contigs have been joined into the same scaffold but the intermittent sequence between them has not been determined.