EPUB Output:Be less aggressive when stripping invalid HTML constructs inserted by MS Word. Fixes regression in the Time recipe.

src/calibre/ebooks/html.py

@@ -331,7 +331,8 @@ class PreProcessor(object):
                   # Convert all entities, since lxml doesn't handle them well
                   (re.compile(r'&(\S+?);'), convert_entities),
                   # Remove the <![if/endif tags inserted by everybody's darling, MS Word
-                  (re.compile(r'(?i)<{0,1}!\[(end){0,1}if[^>]*>'), lambda match: ''),
+                  (re.compile(r'</{0,1}!\[(end){0,1}if\]{0,1}>', re.IGNORECASE), 
+                   lambda match: ''),
                   ]
                      
     # Fix pdftohtml markup