There are a couple of facets to this:
1) speed from knowing shortcuts (keyboard shortcuts, -- coercers, etc.)
2) speed from knowing the right resources/methods that are available (e.g. using a pivot table as opposed to trying to recreate the same thing manually)
3) speed from experience and not having to look up (or think about) syntax (e.g. what's the third argument of a VLOOKUP?, no longer forgetting to put an End If at the appropriate place in code and having to debug, etc.)
4) speed from experience about designing spreadsheets well (particularly with future iterations -- not hardcoding data, allowing for/anticipating expansion, etc.)
5) speed from experience about knowing which function or VBA method is the best way to solve your problem (ties in with #4 above)
6) speed from general typing skills, which will (of course) be a completely separate consideration from any Excel/spreadsheet knowledge
7+) others go here.
So, it depends on how you define the "average" user, and it depends on the knowledge a person will learn in your class; and, of course, it depends on the complexity of the task. A "novice" can probably add A1 and A2 about as fast as an "expert", but a task that takes a "novice" several weeks to do can probably be done by an "expert" in less than a day.
With that said, if I had to make up a number, I'd say 10x as fast when you factor in keyboard shortcuts (particularly not using the mouse) is certainly reasonable (what takes a novice 2 weeks takes him/her 1 day).
Amazingly fast people (the types who are so ridiculously fast in Excel that onlookers can't even see the spreadsheet when s/he is working--only flashing on the screen) maybe 20x+ (a month's work for a novice in one day for an expert). Then again, it's a very real possibility that an "expert" can fully automate tasks that completely eliminate the need for the "novice" at all. In that case, 0 labor hours compared with a full time employee is well more than 20x