Why is the magnetometer data not used when calculating the heading angle?

Why is the magnetometer data not used when calculating the heading angle?

Because it it prone to interference and heading jumps when correcting itself. The BNO085 data sheet shows less drift in Game rotation vector mode (no mag).