CAMEFF P36B — Mc Histogram Equivalence Fix

P36B corrects the real-catalog magnitude-completeness parity defect identified by P36A.

Root cause

The former C17 code derived a bin using division by 0.1 and an epsilon. NumPy's arange derives an effective binary64 step from:

(start + step) - start

and then creates explicit edges. These behaviors differ at decimal boundaries.

Validation

• C17 build and native unit test: PASS
• Frozen 36-case equivalence suite: 36/36 PASS
• Real-catalog Mc parity: 4/4

The complete ETAS real-catalog parity target is included. Run it locally before creating the corrective release.

No FM-v1.4 mathematics was changed.