This doesn't deal with duplicates. If you want to deal with duplicates, I think you change it to:

sub nn {
    @_{@_}=0;@_=sort{$a<=>$b}keys%_;
    $_[1]-$_[0]>$_[$_]-$_[$_-1]and@_[0,1]=@_[$_,$_-1]for 2..$#_;
    @_[0,1]
}
[download]

That brings the total from 86 to 99. I'm not sure if the spec requires that duplicates be treated as one element or separate elements.