gellyfish's code is the way to solve this problem, but I'll point out a couple mistakes in your code to help you avoid them in the future.

print F2 <TAG>;
You want to put TAG in quotes. The line should be print F2 "<TAG>";

This is something that using warnings would have helped you catch. Your program is attempting to read a line from the filehandle TAG, and print that to filehandle F2.

# ./test.pl
Name "main::TAG" used only once: possible typo at ./test.pl line 9.
readline() on unopened filehandle TAG at ./test.pl line 9.
readline() on unopened filehandle TAG at ./test.pl line 9.
readline() on unopened filehandle TAG at ./test.pl line 9.
readline() on unopened filehandle TAG at ./test.pl line 9.
[download]

Also, although it isn't broken in your example, your regular expression can use some work. You are matching anything that isn't ">", and then using it in a numerical comparison. This will be a problem if you capture something non-numeric. A better idea would be to write your regex like this:

if($file =~ /<ins cnt="(\d+)">/) {
  for ($x=0; $x < $1; $x++) {
    print F2 "<TAG>";
  }
}
[download]

What I changed:

• Matching on digits only for the count.
• If $line doesn't match the pattern, your code doesn't attempt to use $1 in a for loop.
• I removed the g modified from your regex, as I don't think you intended for it in this case.
• Nitpicking: I changed the for loop condition from $x != $1. It does not matter for this example, but it's less likely to get caught in an infinite loop when you are doing more complex things (like possibly modifiying $x inside your loop. Note that there are more Perlish ways of writing this, including print F2 "<TAG>" for(1 .. $1).