As others have pointed out, the new lines in the original source header strings are the culprit.

What is generally missed about header-based extraction with HTML::TableExtract, however, is that the strings that you use to define the headers will eventually be turned into case-insensitive regular expressions.

So change the following part:

my $te = new HTML::TableExtract( headers => ["Product Name",
                                             "Software Version",
                                             "Compatible with Versions
+",
                                             "Date" ] );
[download]

to this (notice the single quotes..otherwise you'll have to escape your backslashes):

my $te = new HTML::TableExtract( headers => ['Product\s+Name',
                                             'Software\s+Version',
                                             'Compatible\s+with\s+Vers
+ions",
                                             'Date' ] );
[download]

...and things will work as you expect. Also note that rather than strings, you can pass pre-compiled regexps from qr//, like so:

my $te = new HTML::TableExtract( headers => [qr/Product\s+Name/,
                                             qr/Software\s+Version/,
                                             qr/Compatible\s+with\s+Ve
+rsions/,
                                             'Date' ] );
[download]

Cheers,
Matt