This is the source of the xml_compare sub:

sub xml_compare {
# ============================================================

=item * xml_compare( $xml1, $xml2 ) - Compares xml for content

Compares two dumped Perl data structures (that is, compares the xml) f
+or
identity in content. Use this function rather than perl's built-in str
+ing 
comparison, especially when dealing with perl data that is memory-loca
+tion 
dependent (which pretty much means all references).  This function wil
+l 

return true for any two perl data that are either deep clones of each 
other, or identical. This method is exported by default.

=cut

# ------------------------------------------------------------
    my $xml1 = shift;
    my $xml2 = shift;

    $xml1 =~ s/(<[^>]*)\smemory_address="\dx[A-Za-z0-9]+"([^<]*>)/$1$2
+/g;
    $xml2 =~ s/(<[^>]*)\smemory_address="\dx[A-Za-z0-9]+"([^<]*>)/$1$2
+/g;
    $xml1 =~ s/(<[^>]*)\sdefined=\"false\"([^<]>)/$1$2/g; # For backwa
+rds 
    $xml2 =~ s/(<[^>]*)\sdefined=\"false\"([^<]>)/$1$2/g; # compatibil
+ity
    $xml1 =~ s/<\?xml .*>//; # Ignore XML declaration
    $xml2 =~ s/<\?xml .*>//;
    $xml1 =~ s/<\!DOCTYPE perldata \[.*\]>//s; # Remove DTD
    $xml2 =~ s/<\!DOCTYPE perldata \[.*\]>//s;
    $xml1 =~ s/^\n//gm; # Remove empty newlines
    $xml2 =~ s/^\n//gm;

    return not( $xml1 cmp $xml2 );
}
[download]

What do you see if you compare $rexmlified and $ssxmlified yourself? They must look like simple textual xml don't they? You could try runnung it through the substitutions above and see where they differ.