What Am I missing?

If you ask me, it depends on what you're trying to learn.

Consider

 $ perl -MData::Dump -e " my@f=1..3;my@g=4..6;push@f,\@g;push@g,@f;dd\
+@f,\@g; "
do {
  my $a = [1, 2, 3, [4, 5, 6, 1, 2, 3, 'fix']];
  $a->[3][6] = $a->[3];
  ($a, $a->[3]);
}
[download]

Do you notice 'fix'? Do you notice how the self-referential-ness is established in a second step?

Employ Basic debugging checklist and compare that to your Dumper output

@data1 = (
       'one',
       'won',
       [
         'two',
         'too',
         'to',
         \@data1
       ]
     );
@data2 = @{$data1[2]};
[download]

How is it different from what you originally had? Different from my dd output?

The order of operations isn't the same, @data2 is set to the values of some anonymous array, as opposed to a reference to @data2 being pushed onto @data1 -- the two references will never be the same

Could there be an error in the book?

I wouldn't know, but I don't quite see the lesson its trying to teach -- DD isn't the greatest serialization tool, and string-eval isn't the greatest way to Undumper / Data::Undump

or perl?

Well no, but you could argue it's a bug in Data::Dumper-s prefix/varname feature