53 objects in GP_MAIN_GAME_E carry a Phase_1 record, in TWO families: 29 are the resource manifest the corpus already owns (Phase_N = 4 fields) and 24 are the settings table (Phase_N = 31-90 fields). That answers the 24-vs-28 puzzle left open by the scoring entry -- the 29 is a different table, not the settings. Camera: three chase rigs in metres, 13 of 14 fields identical in every stage -- Nose (0, 4.5, 7), Near (0, 10, 40), Far (0, 15, 80), FOV 0.92; only CameraFar varies, once. Player: BulletLimit 512 / HomingLimit 256 / LaserLimit 32 and the three 0.30 axis adjustments are constant, GravityFactor is non-zero in 4 of 24 stages, and IsBoss16Enable appears in exactly ONE object -- the first per-stage handle for a family whose filenames do not resolve. Difficulty_Easy/Normal/Hard is a SECOND difficulty record (8 damage and guidance multipliers), separate from Score_*. New doc structures/stage-settings-table.md; mission_scoring.py extended. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PMRJjbxLqZtsb5Vb7KunPE