Even though it sounds like homework (summer school?) ... here's a general solution for any number of arrays:

sub merge {
  my %h;
  for my $l (0..$#_) {
    $h{$_}[$l] = $_ for @{$_[$l]};
  }
  map [ map { defined $_ ? $_ : '-' } @{$h{$_}}[0..$#_] ],
      sort { $a <=> $b } keys %h;
}

### Example ###

use strict;
use Data::Dumper;

my @h = merge [3,5,2,6], [7,3,6], [1,6,2,9], [0,3,4,1];
print Dumper \@h;
[download]

Your code could be cleaner, but algorithmically, you've got the right idea.

   MeowChow                                   
               s aamecha.s a..a\u$&owag.print