There's a distinction between nonexistent and nonexistent; if you pass array elements into to a sub, even nonexistent slots will be created and assigned, and the underlying array expanded as appropriate, whereas if you pass in an array, it is flattened into a list, and nonexisting array slots aren't made into aliases.

The behaviour might be called a bug or a malfunction, but thinking about it - it could be taken as a feature also, e.g. for chaining subs where you pass @_ along, as in

#!/usr/bin/perl
$\ = "\n";
@y[3,5] = (13,15);
@x = x(@y);

print '@y = (',join (',', map {"'$_'"} @y),')';
print '@x = (',join (',', map {"'$_'"} @x),')';

sub x {
    $_++ for @_;
    @_[1,4] = (2,5);
    &z;
}
sub z {
    $_ *= 3 for @_;
    @_;
}
__END__
@y = ('','','','42','','48')
@x = ('3','6','3','42','15','48')
[download]

and expect non-existing slots of the original array not to be modified, i.e. you want the holes preserved. I guess that could make perfect sense for e.g. matrix operations, where you don't want undefined elements of sparse arrays spring into existense merely because you $_ **= 2 for @_.

Of course that behaviour should be documented somewhere (not only in PM threads :-)

_($_=" "x(1<<5)."?\n".q·/)Oo.  G°\        /
                              /\_¯/(q    /
----------------------------  \__(m.====·.(_("always off the crowd"))."·
");sub _{s./.($e="'Itrs `mnsgdq Gdbj O`qkdq")=~y/"-y/#-z/;$e.e && print}