does your data always look like pseudo-HTML? tagged? maybe you can have better results with HTML::TokeParser

here's some untested code:

my $p = HTML::TokeParser->new($html) || die "Can't tokenize: $!";
# get each <tag1> alone.
while (my $token = $p->get_tag('tag1')) {
    # store the original text
    $origtext = $token->[3];

    # get data between <MYTAG></MYTAG>
    my $myTag = $p->get_tag('MYTAG');
    my $text = $p->get_text('/MYTAG');
    
    if ($text ne '') {
        
        # tag is not empty.. so $origtext retains
        # the data we want..

        # ... do whatever with $origtext and move on
    }
}
[download]

Instead of using a multine regex, you could consider looping over the lines in turn and removing every six lines where there's a match on the fourth.

I think the code is pretty self-explanatory (@lines contains your file):

my $i = 3;                             # from line 4 ...
while ($i < @lines - 2) {              # ... until last but 3
    if ($lines[$i] =~ m!<MYTAG></MYTAG>!) {
        splice(@lines, $i - 3, 6);     # kill 6 lines from $i - 3
    } else {
        $i++;
    }
}
[download]

— Arien