Decomposing a regex into component parts makes conceptualization and maintenance easier.

perl -wMstrict -le
"my $beginning  = qr{ \A FOO }xms;
 my $not_ending = qr{ (?! (?: BAR | BAZ) \z ) }xms;
 my $ending     = qr{ .*                 \z   }xms;
 for my $string (@ARGV) {
  if ($string =~ m{ $beginning $not_ending ($ending) }xms) {
   print qq($string accepted ('$1' to right))
   }
  else {
   print qq($string rejected)
   }
  }
" FOOABC FOO_BAR FOOBA FOOBAZZZZ "FOOBAR BAZ" FOOBAR FOOBAZ BARBAZ
FOOABC accepted ('ABC' to right)
FOO_BAR accepted ('_BAR' to right)
FOOBA accepted ('BA' to right)
FOOBAZZZZ accepted ('BAZZZZ' to right)
FOOBAR BAZ accepted ('BAR BAZ' to right)
FOOBAR rejected
FOOBAZ rejected
BARBAZ rejected
[download]

To make your life simpler, avoid as the plague the use of capturing groups in the decomposed regexes (the  qr{ ... } regexes); only use them in the matching regexes (e.g., the  m{ ... }), where it's a lot easier to keep track of them.