新六章有 12 节读错公开包字段，整节只剩未判定，而且六章绕过了口径替换。按资料包改读路径，先拼六章再统一替换措辞，并把中文附录实际丢掉的组合句数出来。

Co-Authored-By: Grok (agent loop)
Co-authored-by: Cursor <cursoragent@cursor.com>

scripts/report_render/appendix_language.py

@@ -368,6 +368,63 @@ FIELD_ZH = {
     "months": "月数",
 }
 
+# Names that already had a Chinese header before the #91 fix.
+# The annual-table filter used to drop every other column. Counts use this set
+# so a later Chinese name still shows up as a column the old filter removed.
+ANNUAL_COLUMN_NAMES_BEFORE_FIX = frozenset(FIELD_ZH)
+
+_APPENDIX_DROPS: dict[str, dict[str, int]] = {
+    "annual_columns": {},
+    "annual_tables": {},
+    "effects": {},
+    "combinations_undetermined": {},
+    "combinations_replaced": {},
+}
+
+
+def reset_appendix_drop_log() -> None:
+    """Clear the per-render drop counts. One Chinese render, then snapshot."""
+    for bucket in _APPENDIX_DROPS.values():
+        bucket.clear()
+
+
+def snapshot_appendix_drop_log() -> dict[str, dict[str, int]]:
+    return {name: dict(bucket) for name, bucket in _APPENDIX_DROPS.items()}
+
+
+def note_dropped_annual_column(field: str) -> None:
+    key = str(field)
+    bucket = _APPENDIX_DROPS["annual_columns"]
+    bucket[key] = bucket.get(key, 0) + 1
+
+
+def note_annual_table() -> None:
+    bucket = _APPENDIX_DROPS["annual_tables"]
+    bucket["seen"] = bucket.get("seen", 0) + 1
+
+
+def _drop_identity(item: dict, text: str) -> str:
+    parts = [
+        str(item.get("rule_id") or ""),
+        str(item.get("source") or ""),
+        str(item.get("bvr_id") or ""),
+        str(item.get("name") or ""),
+        " ".join(str(text).split()),
+    ]
+    return " | ".join(parts)[:240]
+
+
+def _note_effect(item: dict, text: str) -> None:
+    key = _drop_identity(item, text)
+    bucket = _APPENDIX_DROPS["effects"]
+    bucket[key] = bucket.get(key, 0) + 1
+
+
+def _note_combination(item: dict, text: str, kind: str) -> None:
+    key = _drop_identity(item, text)
+    bucket = _APPENDIX_DROPS[kind]
+    bucket[key] = bucket.get(key, 0) + 1
+
 
 def column_header(field: str, language: str) -> str:
     if language != "zh":
@@ -591,12 +648,19 @@ def chinese_yoga_label(item: dict) -> str:
 def chinese_combination(item: dict) -> str:
     """Use the Chinese combination sentence. An English sentence is not translated word by word."""
     combination = str(item.get("combination") or "").strip()
-    if combination and re.search(r"[\u4e00-\u9fff]", combination) and not _HAN_FUNCTION_RE.search(combination) and not SIGN_EN_RE.search(combination):
+    han = bool(re.search(r"[\u4e00-\u9fff]", combination))
+    function = bool(_HAN_FUNCTION_RE.search(combination)) if combination else False
+    sign = bool(SIGN_EN_RE.search(combination)) if combination else False
+    if combination and han and not function and not sign:
         return combination
     planets = item.get("planets") if isinstance(item.get("planets"), list) else []
     names = [planet_name(planet, "zh") for planet in planets if planet]
     if names:
+        if combination:
+            _note_combination(item, combination, "combinations_replaced")
         return "与".join(names)
+    if combination:
+        _note_combination(item, combination, "combinations_undetermined")
     return "未判定"
 
 
@@ -610,7 +674,15 @@ def yoga_cells(item: dict, language: str) -> list[Any]:
         category_text = "未判定"
     strength = item.get("strength") or "-"
     effects = item.get("effects") if isinstance(item.get("effects"), list) else []
-    kept = [str(effect) for effect in effects if str(effect).strip() and not _has_prose_glue(str(effect))]
+    kept = []
+    for effect in effects:
+        text = str(effect).strip()
+        if not text:
+            continue
+        if _has_prose_glue(text):
+            _note_effect(item, text)
+            continue
+        kept.append(text)
     effects_text = "；".join(kept) or "-"
     return [chinese_yoga_label(item), category_text, strength, chinese_combination(item), effects_text]