You can make this problem easier to solve by switching to HTML::TokeParser::Simple. (Disclaimer: I wrote it)

use strict;

use HTML::TokeParser::Simple;
my $p = HTML::TokeParser::Simple->new(shift||die);

while (my $token = $p->get_token) {
    print $token->return_attr('href'), "\n"
        if $token->is_start_tag('a');
}
[download]

Note that the module is a drop-in replacement for HTML::TokeParser. If you switch to it, your code will function exactly the same and you can simply rewrite the messy bits.