The manner in which Sensor Fusion works, is to get the best out of all nine sensor axes. Every sensor has pros and cons, and sensor fusion is basically using the pros of all the devices, counterbalancing for the cons. For an example, consider the gyroscope. It is going to drift over time, and the drift is going to accumulate, slowly.  By relying on the gyroscope only to find the absolute heading will over time start building up an error. The solution with sensor fusion begins to use the magnetometer as an absolute reference point. If drift is occurring, the system will put more value on the magnetometer and vice-versa. If the magnetometer is erratic, due to some kind of external disturbance or interference, the system will be putting more value on the gyroscope, or the accelerometer to retain an optimal output.