The re-imported FP_XDCR_BMM150 source geometry is correctly scaled:

However, the effective footprint is still incorrect because stale per-pad overrides from my previous attempt remain active and are still being interpreted at 0.001× scale. The imported footprint also specifies a large 0.102 mm solder-mask expansion, producing approximately 0.404 mm mask openings, which should be reviewed for this fine-pitch WLCSP.

I’ll remove every stale pad-level rule, normalize the imported footprint so its effective geometry is 0.200 mm pads on the 0.566 mm array, then set a tighter mask expansion and verify the resulting mask web.

Stale rules are cleared. I’m now applying the corrected source-pad coordinates and a 0.050 mm mask expansion, giving 0.300 mm mask openings and 0.266 mm minimum mask web at 0.566 mm pitch.

Footprint corrected and verified:

All 12 pads now have the expected effective coordinates, circular geometry, and signal-to-ball mapping.