Personally, I would use a different approach by restating the problem: You are interested in all *.mrg files that have no corresponding *.did file :

opendir DIR, $dir
  or die "Couldn't open directory '$dir' : $!";
my @files = grep { /(.*)\.mrg$/ and not -f "$dir/$1.did" } readdir DIR
+;
closedir DIR;
[download]

My method might be a bit slower, as for each .mrg file, an additional call to stat will be made, which can be very slow on full directories, but I think that the shorter code makes up for the slower code. If speed should really become an issue, I'd readdir the directories contents into a hash and then check for existence in the hash much like your example:

  opendir DIR, $dir
    or die "Couldn't read '$dir' : $!";
  my @all_files = map { lc $_ } readdir DIR;
  closedir DIR;
  my %did = map { /(.*)\.did$/ and ($1 => 1) } grep { /\.did$/ } @all_
+files;
  my @files = grep { /(.*)\.mrg$/ and not $did{$1} } @all_files;
[download]

Also, I don't think that production code should contain references to Perlmonks node IDs, but rather an explanation of what happens :

  # return a list or a reference to an array, depending
  # on what the caller wants:
  return wantarray ? @files : \@files;
[download]

Update: Added "faster" alternative

Update 2: Fixed code in response to merlyns bugfinding

perl -MHTTP::Daemon -MHTTP::Response -MLWP::Simple -e ' ;    # The  
$d = new HTTP::Daemon and fork and getprint $d->url and exit;#spider
($c = $d->accept())->get_request(); $c->send_response( new   #in the
HTTP::Response(200,$_,$_,qq(Just another Perl hacker\n))); ' #  web
[download]