...but come to think of it, it is simpler to just fix your original version:

sub append_n
{
      my ( $str, $array ) = @_;
      my $nstring = "N" x length($str);

      foreach my $sbstr ( @$array )
      {
        my $ofs = 0;
        while ( ( my $pos = index $str, $sbstr, $ofs ) > -1 ) {
          substr ($nstring, $pos, length ($sbstr)) = $sbstr;
          $ofs = $pos + 1;
        }
      }
      return $nstring;
}
__END__
GATTACGNNNNGCGCTCGNNNAACGGCA
GATTACGAGNNNNNNNCGTGTAANNNNN
NNNTACGAGTGGCGCTCGTGNNNNNNNN
NNNNNNNNNNGGNNNNNNNNNNNNGGNN
[download]

...and about 3X faster too:

          Rate bitvec substr
bitvec 18618/s     --   -74%
substr 70275/s   277%     --
[download]

In hindsight this is not surprising, since my (vec-based) solution does everything that yours does plus the unnecessary bit-vector stuff. Doh!