makes sense to me:

[^>] - matches 1
\d+ - matches 2
[a-z]* - matches empty string
[^<] - matches a
[download]

And it's the first place that you have digits. What are you thinking it should have matched?

Also, generally, you don't want to use regex against angle-bracket stuff. Use something based on HTML::Parser, XML::Parser, or XML::LibXML, please.