diff --git a/src/fenigma/ocr.py b/src/fenigma/ocr.py
index 5c98e2e..aa27995 100644
--- a/src/fenigma/ocr.py
+++ b/src/fenigma/ocr.py
@@ -609,6 +609,11 @@ def parse_train_intel(text: str) -> list[dict]:
 _DESTROYED_RE = re.compile(
     rf"([A-Za-z]+)[^A-Za-z0-9\s]{{1,2}}({_DIGIT_CLASS}+)\s*Destroyed", re.IGNORECASE
 )
+# Same digit-vs-letter-id split as the header regexes: an ad-hoc "Enemy
+# X Y" installation's id (after squash_enemy_names()) is letters, not
+# digits, so it needs its own pattern, _DESTROYED_RE's digit class won't
+# match it at all ("Enemy#SignalStation Destroyed" was silently dropped).
+_ENEMY_DESTROYED_RE = re.compile(r"(Enemy)#([A-Za-z]+)\s*Destroyed", re.IGNORECASE)
 
 
 def parse_destroyed(text: str) -> set[tuple[TargetType, str]]:
@@ -619,6 +624,12 @@ def parse_destroyed(text: str) -> set[tuple[TargetType, str]]:
         if target_type is None:
             continue
         destroyed.add((target_type, _fix_id_digits(num)))
+    for m in _ENEMY_DESTROYED_RE.finditer(text):
+        type_word, letter_id = m.groups()
+        target_type = _resolve_target_type(type_word)
+        if target_type is None:
+            continue
+        destroyed.add((target_type, letter_id))
     return destroyed