I can't say much without having examples of data which produce these "erronious results," but I've two guesses:

1. $in{for} is empty. This means that the regex is using whatever the last successful match or replacement was. This is probably not what you want. Postfix $in{for} with .* to avoid this.
2. $in{for} contains metacharacters. That is, things like . and * which have special meaning to a regular expression. To get around this, use \Q$in{for}\E instead of $in{for} in the regex; or, somewhere above the regex, run $in{for} through quotemeta.

perl -pe '"I lo*`+$^X$\"$]!$/"=~m%(.*)%s;$_=$1;y^`+*^e v^#$&V"+@( NO CARRIER'

That's hard to say without seeing some representative data. Show us a value for $in{for}, and @thread[0,4] pair that you expect to match (but doesn't), and a pair that does match that you expect should not.

Ok, here's (I hope) a little clarification. here's a big snipet of code

my @dat_files = <$board_dir/*.dat>;
$q = 0;
foreach (@dat_files) {
    $number;
    $number = $_;
    $number =~ s/\/var\/www\/cgi-bin\/2930forum\/data\///g;
    $number =~ s/\.dat//g;
    open THREAD, "$_" or die "Can't open .dat file: $!";
    $x = 0;
    while (<THREAD>) {
        $thread_data[$x] = $_;
         $x++;
     }
     close THREAD;
     foreach (@thread_data) {
         @details = split /\|/, $_;
         if ($details[4] =~ m/\Q$in{for}\E/i) {
             $found[$q] = $number;
              $q++;
         }
    }
}
[download]

It's a little sploppy at this point, but I'm just trying to get valid results at this point, I'll clean it up later.

$in{for}: is defined my form input (I've been using simple searches like "cheese")
$thread_data[4]: is the messages posted in every thread. I could potentially contain just about anything except empty.
$thread_data[0]: I just realized isn't used. Insted it's $number. Which is just a number string denoting the thread number

I do the search and get results. Some of the threads contain the searchword $in{for} and others do not. For example I put in $in{for} = cheese and get around 30 results containing messages like:
Hard work pays off after time, but lazyness always pays off now.

One thing I just noticed is that many of the results are in numerical order. for example I get results like 121, 122, 124, 125, 126, 127, 128, 129, 21, 282 ,321, 343, 344, 345, etc

This code has a couple of problems. One is that the handling of filenames, particularly the method of extracting a number from them, is highly suspect. Once you think you've extracted $number, try printing both the full filename and $number.

It looks like you're trying to accumulate a list of thread nubmers that contain matches. Since you say that $thread_data[0] is the same as $number, this might be easier like so:

my @dat_files = <$bboard/*.dat>;
my %found = ();
foreach my $file ( @dat_files ) {
    open(DAT, $file) or die "$file: $!";
    while ( <DAT> ) {
        my @thread_data = split "|";
        if ( $thread_data[4] =~ m/\Q$in{$for}\E/i ) {
            $found{$thread_data[0]}++;
        }
    }
    close(DAT);
}
[download]

The keys of %found are now the thread numbers taht contain a match, and the corresponding values are the number of matches.

my $IN = quotemeta($in);
if ($thread_data[4] =~ m/$IN{for}/i) {
      print $thread_data[0];
}
[download]

I can't see nothing wrong in the snipet... Maybe the problem lays in how the @thread_data array is filled? are you shure that when you test the next thread_data _all_ @thread_data elements have changed?