I read it the same way you do.

use strict;
use warnings;

use Test::More qw( no_plan );

my $ch = chr(0xE9);  # lowercase e acute

# Byte semantics
utf8::downgrade($ch);
ok( $ch !~ /\w/ );
ok( $ch =~ /\W/ );
ok( $ch !~ /[\w]/ );
ok( $ch =~ /[\W]/ );

# Unicode semantics
utf8::upgrade($ch);
ok( $ch =~ /\w/ );
ok( $ch !~ /\W/ );
ok( $ch =~ /[\w]/ );   # Should fail according to the docs.
ok( $ch !~ /[\W]/ );   # Should fail according to the docs.
[download]

[\w] seems to work correctly (as in contrary to the docs), but not [\W].

>c:\progs\perl588\bin\perl test.pl
ok 1
ok 2
ok 3
ok 4
ok 5
ok 6
ok 7
not ok 8
#   Failed test in test.pl at line 20.
1..8
# Looks like you failed 1 test of 8.
[download]

Same in 5.10.0.

>c:\progs\perl5100\bin\perl test.pl
ok 1
ok 2
ok 3
ok 4
ok 5
ok 6
ok 7
not ok 8
#   Failed test at test.pl line 20.
1..8
# Looks like you failed 1 test of 8.
[download]

Although the bit you quoted was removed from from the docs.

Update: It might help if I actually used a character class in my tests. Fixed.