Ferrite core will change the distortion of the magnetic field so if you add or remove it then I would expect you would need to perform mag calibration. But if it's working - don't change it
[edit] There are a couple of bugs in the latest (v4) 9150 test sketch which may be effecting you.
The start-up calibration can product odd results. Manifests as magnetometer struggling to keep yaw drift in check (head wobbles a bit). Workaround is to press the button again.
The 'recentre view' code doesn't clear magnetometer readings so pressing multiple times results in a build up of mag compensation errors.
The code which calculates the transformation matrix which "un-distorts" the magnetometer readings can fail if the raw data is too far off centre ( the red sphere is way off).
All are fixed in the production release

(... soon)
Rob