Leave it in place, but ideally figure out how to add the missing pages eventually, if only for the sake of completeness. Sitemaps are only informative. If something is missing from the map, it won't be interpreted as "this isn't intended for crawling." As long as the pages are linked from somewhere, they'll get crawled.
You didn't mention how long you've given this, but note that generating a sitemap might be computationally expensive if it's a large site, so there might be a delay in regenerating it, either just from the process running, or even because there's a scheduled task that handles it rather than it being done real-time.