ok, it's more C-ish than perl-ish, but i betcha it's faster:

#!/usr/bin/perl -l

my $A = 'ATGGAGTCGACGAATTTGAAGAAT';
my $B = 'xxxxxxATGGAGyxxxTCGAzxxxxCGAATTTGAAxxwGAAT';
my ($An, $Bn) = (0, 0); # relative indices
my ($Ac, $Bc);  # characters
my ($difpos, $difstr) = (0, ''); # keep track of differences

for $Bn (0 .. length($B)-1){
    $Ac = substr($A, $An, 1);
    $Bc = substr($B, $Bn, 1);
    if ($Ac eq $Bc){
        print "$difpos: $difstr" if $difstr;
        $difstr = '';
        $An++;
        $difpos++;
    } else {
        $difstr .= $Bc;
    }
}
[download]

perl -e'$_=q#: 13_2: 12/"{>: 8_4) (_4: 6/2"-2; 3;-2"\2: 5/7\_/\7: 12m m::#;s#:#\n#g;s#(\D)(\d+)#$1x$2#ge;print'