I often find that what starts out as an apparently trivial task of parsing apparantly trivial HTML ends up being very difficult.

Even in this case I would reach for an HTML parser.

#!/usr/bin/perl

use warnings;
use strict;

use HTML::TokeParser::Simple;

my @array = <DATA>;

for my $line (@array){
  
  my $tp = HTML::TokeParser::Simple->new(\$line);
  
  my $cell_data;
  
  while (my $t = $tp->get_token){
    
    $cell_data++, next if 
      $t->is_start_tag('td')       and
      $t->get_attr('height') == 40 and
      $t->get_attr('width')  == 40 and
      $t->get_attr('border') ==  0;
      
    print $t->as_is if $cell_data and $t->is_text;
    
  }

}

__DATA__
<td width='10' height='40' border='0'>cell data 1</td>
<td width='20' height='40' border='0'>cell data 2</td>
<td width='30' height='40' border='0'>cell data 3</td>
<td width='40' height='40' border='0'>cell data 4</td>
<td width='40' height='40' border='0'>cell data 5</td>
<td width='40' height='40' border='0'>cell data 6</td>
<td width='40' height='40' border='0'>cell data 7</td>
<td width='40' height='40' border='0'>cell data 8</td>
[download]

Output:

---------- Capture Output ----------
> "C:\Perl\bin\perl.exe" _new.pl
cell data 4
cell data 5
cell data 6
cell data 7
cell data 8
> Terminated with exit code 0.
[download]

I believe this gives you greater flexibility/adaptability and, for me, is quicker to write. :-)

Hope that helps