The question isn't precise enough...
It depends on the dimension and on the method :
If you measure a distance along Z (cmm axis) with the optics, it's not the same uncertainty than a Z dimension measured by contact, or a X/Y dimension measured by optics or by contact.
In your case, I would try to create a known artifact which represent the measurement, and qualify it with your method (for example, if you have to measure the distance between two holes on a thin part by optics, measure a standard part (not thin) with two holes by contact (on another cmm if you can), and then measure the part by optics, and compare the results...)
I would recommend, if possible, checking <something> with hand tools (micrometer, comparator, etc) and comparing the results to that.
You'll pretty much need to measure once with hand tools, and try a bunch of different methods on the CMM until yo uge a result that works.
The uncertainty shouldn't be an issue if you can deal with the results it spits out
