while (<$fh> ) {
    if ( $_ =~ /$eot/ ) {
     last;
    } else {
        $_ = <$fh>;
        $item = $_;
     $item =~ s/\n//gs;
     $item .= "\n\n";
    }

   if ( $item =~ /$errmsg/ ) {
     push @erritems, $item;
   }
 }
[download]

Maybe I'm just missing something too, and I certainly don't understand the structure of your data file, but here's what I understand of the code fragment above:

• read a line | paragraph;
• test if the line | paragraph has an  $eot match and exit the loop (and error log processing) if so;
• if the line | paragraph is not an  $eot match, throw it away and read the next line | paragraph from the file (update: which is not tested for an  $eot match);
• do some newline massaging on the line | paragraph;
• push the (massaged) line | paragraph to the  @erritems array if it is an  $errmsg match.

What I don't understand is why you're throwing away half the lines | paragraphs in the file up to the point of the  $eot line | paragraph match.

Update: I just noticed you're reading the file in paragraph mode
    $/ = '<table align="left" border="0" cellspacing="0" cellpadding="1"';
and so changed all instances of line(s) to paragraph(s).