performance and the xsl:for-each-group instruction

Does the xsl:for-each-group instruction have a performance benefit over using 
the Muench method, or is it just syntactic candy?  I've been testing 
xsl:for-each-group against the Muench method in Saxon 7.8 and I don't see any 
performance gain.  It would make sense (at least to me) that xsl:for-each-group 
could operate much more efficiently than the Muench method since it knows what 
you want to do is group.