Well, I disconnected my external compass and got the same results you did.
First, in CLI, it reported "Not healthy" as expected.
But like you observed, in the HUD, the HUD's compass still showed some readings, Since the trace cuts the SDA line to the on-board compass, perhaps with no SDA line connected without the external compass hooked up, the CPU picks up a raggedly SDA signal from the on-board compass. No matter. With the external SDA line connected, this stray signal from the on-board SDA should not have an effect. Just my guess.
I would say you cut the trace properly. Have you tried hooking up the external compass and checking the results in CLI and in the HUD? If they are OK, then you're done.
Thanks for testing. Yesterday I rebuild the controller stack of my quadcopter. Installed the external magnetometer again. Did a manual calibration in the Mission Planner and everything is working as it should.
I guess I began to doubt myself because of the compass signal registered in the Flight Data screen without my external compass connected.
Thanks for helping me solve my 'problem'.