From perldoc perlvar:

$-
The number of lines left on the page of the currently selected output channel. Used with formats. (Mnemonic: lines_on_page - lines_printed.)

Punctuation variables are reserved for special purposes in Perl. Also, your second while loop is populating $_, not $-. So if you want the contents of <REPLACE> to be passed through that second substitution operator, you should bind $_ to it, not $-.

See also perldoc perlvar:

$_
The default input and pattern-searching space.

It has a special purpose. You used it correctly in the first while statement, but incorrectly in the second while.

perlvar

$FORMAT_LINES_LEFT
$-
The number of lines left on the page of the currently selected output channel. Used with formats. (Mnemonic: lines_on_page - lines_printed.)

Perl has lots of these special variables. They're independent of each other (though some have related areas of utility.) In the general case, arbitrarily substituting one for another is the wrong thing, and you should expect it to break your code.

It should be $_ like the others explained. Let's look at why it doesn't work with $_. (I don't see how it would work with $- any better.) You're missing parens in your first match. m// returns 1 or undef/() if you have no captures.

{
  my @new;
  $_ = 'bla';
  push @new, $_ =~ m/^bla/;
  print(@new, "\n");  # prints 1
}

{
  my @new;
  $_ = 'bla';
  push @new, $_ =~ m/^(bla)/;
  print(@new, "\n");  # prints bla
}
[download]

$_ =~ is redundant and can simply be ommited. Also, s/...(.*)/...$1/ is the same as s/.../.../. The final product is:

while (<IN>) {
   push @new, m/^(\S+\s+\S+\s+\S+\s+\S+\s+\S+\s+\S+\s+)/;
   print "\n$#num\n";
}

while (<REPLACE>) {
   s/^\S+\s+\S+\s+\S+\s+\S+\s+\S+\s+\S+\s+/$new[$i++]/;
   print OUT $_;
}
[download]