• What is your OS?
• What is your perl version? (perl -v)
• Did you invoke the script with the required -CS command-line option?
  $ perl -CS split2.pl < inputfile

My example was used on UTF-8 encoded files that contained quite a few characters outside of the iso-8895-1 range, so I should have noted the same warnings if my example was seriously flawed.

Is your data secret, or is it sharable, in which case, some of us might want to download it (in a zip) to check.

As you converted my command-line example to a script, maybe it would be a goor idea to show what the script looks like. You might have missed a crucial issue. It might look a bit like this:

use strict;
use warnings;
use autodie;

local $/ = \3000;
my $i = "0000";
while (<>) {
    my $fn = "zz" . $i++;
    open my $fh, ">:encoding(utf-8)", $fn or die "$fn: $!";
    print $fh $_;
    close $fh;
    }
[download]

OS: Windows 10 Home
perl 5, version 14, subversion 2 (v5.14.2) built for MSWin32-x86-multi-thread

This is my script split2.pl


  
use strict;
use warnings;
use autodie;


$/=\3000;
my$i="000";
while(<>){open my $fh, ">:encoding(utf-8)", "input".$i++.".txt";
print $fh $_;
close $fh;}
[download]

If I invoke the script in this way:  perl -CS split2.pl <input.txt
I obtain this message

 utf8 "\xE1" does not map to Unicode at split2.pl line 11, <> chunk 2.
Close with partial character at (eval 21) line 67, <> chunk 2.
[download]

and only the first fragment is created "input000.txt"

If I run the script without -CS, no warning message and all the files are created. But they include inintelligible characters and not my greek text splitted.

I can share my greek text (346 kB) but I do not exactly in which way I can do from here.

-CS uses UTF-8 for standard input and output, but the diamond operator uses the ARGV handle, not STDIN.

Use -Ci to set UTF-8 encoding of all input; or use -CD to use utf-8 for all input and output, and you can even drop the encoding from the open line.

Update: Using redirection works for me (Linux). Are you sure the input is encoded in UTF-8?

Update2: Verified with the file you linked to.

perl -CS split2.pl < input.txt
[download]

works correctly on Linux.
map{substr$_->[0],$_->[1]||0,1}[\*||{},3],[[]],[ref qr-1,-,-1],[{}],[sub{}^*ARGV,3]

This is my input file

https://ufile.io/v0g1c