The only valid argument for placing the magnet (and sensor) closer to the hub is that the linear speed of the magnet relative to the sensor is lower so that there is less chance that a revolution will be missed (not counted). There has been one actual instance in my experience where moving the magnet and sensor closer to the hub made a difference in the reliability of a cycling computer, so I do this now routinely. The magnet is so light in comparison to, say, a valve stem, that the actual placement probably won't make a detectable difference in how the wheel rolls.
-soma5