The last 'Ã' looks very suspicious: latin1 characters outside of the basic 0-127 range are stored on 2 bytes in UTF-8, they look like 'Ã?', a lone 'Ã' is certainly an error. My guess would be that a cut'n paste went wrong and the last character of the string was lost.

I used this bit of code to check the string BTW:

#!/usr/bin/perl -w
use strict;

use XML::Parser;
use Text::Iconv;

my $text= <DATA>;

Text::Iconv->raise_error(1);
my $converter= Text::Iconv->new( utf8 => 'latin1');
my $in_latin1=  $converter->convert( $text);
print "text in latin1: $in_latin1\n";

__DATA__
<title>Os cem melhores contos brasileiros do século /Italo  Moriconi,
+ organização, introdução e refer</title>
[download]