The screw_undetermined rule compared the axial rows of our answer with those of
the reference as deposited. Which axis carries the disagreement is a property of
the setting, so where the two cells are a permutation of each other the rule
looked at one axis and SPACE_GROUP_SCREW_UNDETERMINED named another, and a row
whose open screw the run had explicitly declined to decide was scored as a wrong
space group. The comparison now runs against the offered alternative of the
reference's number, which is already written in our axes; a group offered under
more than one setting fires if any of them fits.
Verdicts moved on 7mzt only (fail/sym_screw -> unscored/screw_undetermined) in
both persisted runs re-scored; no other row changed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013nW6FNRP1bBJJ8pfHiByAT
leonarski_fandClaude Opus 5
ae19223549