Maybe something like this:

#!/usr/bin/perl -wl
use strict;

while (<DATA>) {
    my $wanted;
    if (($wanted) = $_ =~ m/^gif"><\/td><td>(?:<b>)?([A-Z0-9\(\)\[\]]+
+?)<\/td>/) {
        print $wanted;
    }
}


__DATA__
gif"></td><td>XXX</td></tr>
gif"></td><td>XXX</td></tr>
gif"></td><td><b>XXX</td><td><b>yyy</td></tr>
[download]

Notes:

(?:<b>)?
[download]

- makes the <b> optional, and doesn't capture it

([A-Z0-9\(\)\[\]]+?)
[download]

- is where you capture your wanted string. Note the trailing +? - this makes it non-greedy, so that it only captures up until the first </td>

Having said the above, be aware that there are several CPAN modules available for parsing HTML.

Hope this helps,
Darren :)