Independently I have arrived at a similar solution.

#!/usr/bin/env perl 
use strict;
use warnings;

my %excluded = map { $_ => 1 }
    qw( a about although also an and another are as at be been
    before between but by can do during for from has how however in in
+to is
    it many may more most etc );
my %count;

{
    local $/ = "";
    while (<>) {
        tr {A-Z':@~,.()?*%/[]="-}{a-z}d;
        foreach (split) {
            $count{$_}++ unless $excluded{$_};
        }
    }
}

foreach my $word (sort { $count{$a} <=> $count{$b} or $a cmp $b } keys
+ %count) {
    print "$count{$word} $word\n";
}
[download]

I've leveraged the requirement of only lowercasing the ascii letters by incorporating it into the tr/// and I've gone for paragraph mode instead of a single slurp, just in case :-)

Both solutions run in similar times and about 100x faster than the original code:

$ time ./11116620.pl < Frankenstein.txt > orig.out

real    0m9.381s
user    0m9.366s
sys    0m0.007s
$ time ./wordcount.pl < Frankenstein.txt > hippo.out

real    0m0.089s
user    0m0.081s
sys    0m0.008s
$ time ./tybalt.pl < Frankenstein.txt > tybalt.out

real    0m0.090s
user    0m0.084s
sys    0m0.005s
[download]

There are some minor differences between all three outputs but without a tighter spec these aren't overly concerning.