because digital signals are not signals in the classic sense, they're simply a serial stream of binary numbers that need conversion and processing to make any sense at all.
If you simply merged two digi signals, you'd screw up the timing information in both and destroy all information in there.
To do digi mix, you need to dechipher the serial digi streams into a stream of parallel "real" x-bit numbers/values, then multiply these with a constant (mix level), then add the two (making sure that you don't exceed the number of bits wanted), and reduce result to be within your wanted bitdepth.
A simple addition of "real" x-bit values wouldn't do either, as two 16-bit figures would on average add up to a 17-bit number - so some data reduction is necessary, e.g. removal of lsb.
It's fair to say that digi mixing is far, far more complicated than it's nalog counterpart. Which is also the reason for the huge difference between analouge and digital mix quality.
Jakob E.