This is better handled with XPath.

m( <pg [^>]* > # A <pg> tag with optional body
   ( # Capture START
       (?:
           # Skip past all the non-tag-beginng-like things but be sure
           # not to allow backtracking.
           (?> [^<]+ )
           | # OR
           # Match something like a tag beginning as long as it
           # isn't an opening for <pg
           < (?! pg\s )
       )
   ) # Capture END
   </pg> )x;
[download]