When I run mongorestore the first time into a fresh database, it's significantly faster than when I run it the second time (when all data is already present).
I'd expect the run with dupes to be faster since less data is actually being written.