I assume that the OP is talking about framing and not focus. The accuracy of the viewfinder of an SLR is determined by the size of the mirror. There are several competing problems here. If the mirror is designed to provide 100% coverage the inertia of the mirror becomes a problem. Camera shake also increases with mirror size. Most SLR's are therefore a compromize and provide less than 100% coverage. Rangefinders with frame lines actually show more than 100% coverage for most lenses. I think that most people do not really think that somewhat less than full coverage to be a problem.
Rangefinder parallax practically guarantees you are not getting what your framelines show.
For example, subjects at infinity do not move when you focus closer. But the framelines (of a parallax-corrected rangefinder) do move. The framelines move to indicate boundaries of subject at the plane of focus.
This problem only crops up if you are considering including a sliver of sky and mountaintop in your composition, then you should "remember" where the framelines were when the lens was set at infinity.
I do this somewhat subconsciously when I shoot rangefinders and it's not as hard as it sounds.
I don't think it's about mirror size only, or even mainly. The increase in area, thus mass, of a mirror which reflects 95% of the frame area, to one which reflects 100% is slightly more than 5%. I think it's much more about precise alignment. If a viewfinder shows less than 100%, the viewfinder image can "float" within the actual frame, which is borne out by tests done on cameras with less than full coverage which often show the viewfinder image to not be centered. A 100% viewfinder must be very accurate to be useful, and that is harder to do and more expensive.
Originally Posted by Gerald C Koch
Some quiet, low vibration cameras have nearly 100% viewfinder images. The OM-1 was one of the first to address mirror noise and vibration in a determined way, and its coverage is something like 97%.
One extra thing also - different Leica M cameras have different accuracy, and it depends also from lens you use.
M6 (0.72 vinder version) is very accurate on 50mm on close focus, but very wrong on infinity showing only around 85% of what you get on film. For 35mm lenses M6 is the best - biggest amount of viewfinder is in use, and very precise on close and infinity.
M3 is very good on infinity, but on close focus you get less on film than in viewfinder.
For other M bodies I can't say anything since I don't have them. I have made this conclusions by testing them in parallel with nikon F3 who is for me prefect accurate machine :). Maybe someone else had made some testing for M2, M4? I guess M7 and MP are the same as my M6.