Adds the lines's words don't have yet, if it's already been read at modified (a frame read in the viewer, from a video the indexer sampled); an item not read yet is left for its full read. Returns how many lines were new.