Maybe now is a good time to show the exact code you are running and the exact input (again, as hexdump) you are giving it, and also to describe what method you are using to inspect the output.

For me, on Perl 5.20, on Windows 7, with the Latin-1 codepage, I get the following output from the program I posted with the input file, which shows some more "characters" on output, but that is expected because my terminal is not set to UTF-8:

Ruler : [12345678901234567890]
Input : [YearÔÖÑJEDocSrcP_USE_DATE  P_DATE  CurLine]
20    : [YearÔÖÑJEDocSrcP_USE_D]
[download]

On Perl 5.20, on Windows 7, with the UTF-8 codepage (via chcp 65001), I get the following output from the program I posted with the input file, which has 20 characters (not bytes) on output, as I expect:

The script I'm running is:

#!/usr/bin/perl -w
use strict;
use Encode qw/encode decode/;

open (INFILE, "<:encoding(UTF-8)", "utf8.txt") || die "blah blah blah"
+;
open (OUTFILE, ">:encoding(UTF-8)", "oututf8.txt") || die "blah blah";
binmode STDOUT, ':encoding(UTF-8)';

print "Ruler : [12345678901234567890]\n";
while (my $line = <INFILE>) {
  chomp ($line);
  $line =~ s!^\N{BYTE ORDER MARK}!!;
  print "Input : [$line]\n";
  my $linestart = substr($line,0,20);
  my $outline = $linestart;
  print "20    : [$outline]\n";
  print "---\n";
  print OUTFILE "$outline\n";
}
close (INFILE);
[download]