Using individual named vars makes this awkward. Using a hash, makes it simple:

#! perl =slw
use strict;
use Data::Dump qw[ pp ];

my $seq1="AAAGATCGTG";
my $seq2="AGTACTTTCC";

my %subs;
for my $p ( 0 .. length( $seq1 ) - 1 ) {
    my $c1 = substr $seq1, $p, 1;
    my $c2 = substr $seq2, $p, 1;
    next if $c1 eq $c2;
    ++$subs{ $c1 . '2' . $c2 };
}

pp \%subs;

__END__
C:\test>junk.pl
{ A2C => 1, A2G => 1, A2T => 1, C2T => 1, G2A => 1, G2C => 1, G2T => 1
+, T2C => 1 }
[download]