In your first test, you're splitting on 'comma', and your input string conveniently contains commas. So you get the desired behavior of @input receiving elements ab, cd, efghi, and jkl.

In your second test, you're splitting on @, which is not present in your input string at all. Thus (contrary to what your #comment says), the string isn't being split; it doesn't contain anything that matches your split pattern. Consequently, you get the undesired behavior of @input receiving a single element, "ab,cd,efghi,jkl"

If you're dealing with CSV files, or anything more complex than very simple un-quoted comma delemited text, Text::CSV will save you headaches and development time in the longrun.

Is there anything special that I haven't read about using @?

Others have pointed out the conceptual problem illustrated by the example code in the OP. One further point about  @ is that it is the sigil of arrays in Perl, e.g.  @array etc. Arrays interpolate into double-quoted strings and into regexes just as scalars do. If you had been trying to split on a string containing an '@' character that might have been mistaken by the compiler for an array name, it would have been necessary to escape the '@' character (see example below). The naked '@' in the  /@/ regex cannot be mistaken for an array name and so needs no escape.

Try this code without the \ escape (and with warnings and strictures enabled! – see note below):

>perl -wMstrict -le
"my $str = 'XXX@okYYY';
 my @pieces = split /\@ok/, $str;
 printf qq{'$_' } for @pieces;
"
'XXX' 'YYY'
[download]

NOTE: As you are learning Perl, it is highly advisable to enable warnings and strictures (warnings and strict) in your code. The diagnostics pragma is also often very useful to learners. The incantation
    use warnings;
    use strict;
    use diagnostics;
can be wonderfully informative, and I and even quite experienced Perlers routinely enable warnings and strictures.

As noted, splitting on anything that's NOT in the string results in passing the WHOLE string direct to the array.

In any case, perldoc -f split may be a starting point ... or perldoc perlretut.

print "$input[$len-1]"; # splits properly into 4

You can just change it to simply:

print "@input";

---output---