I mean, surely if you are doing something that requires this level of precision, you could just ask the user to input its current known location? I doubt that even if the user misdialed by ten or twenty meters the difference in compensation would matter (or even if the camera was actually moving around).