said by OZO:
Cell phones have intrinsic and big latency ...
I am not sure that there is any intrinsic high latency in cellular communication. All EM waves travel at the speed of light. Given a fixed speed of information travel, the time it takes to go from point a to b is purely a function of distance.
Note that this distance is not geomtric distance, but an effective distance. In other words, latency can be large if the signals are improperly routed. However, this is not a fundamental, physical constraint.
I expect that spectrum (as you noted) and transmission losses are the discrminating factors in voice quality. Dissimilar codecs would require an extra decoding - recoding step, perhaps contributing to poor quality.
To the OP, have you tried forcing the gsm codec? The narrow band codecs should not degrade clarity ( in my opinion, but this is a very subjective issue), but would make the audio sound less "natural".
This is partly an educated guess, so I would really appreciate if you report back your findings!