Put another way, when I'm actually working on a project in the language, what's the likelihood that someone who I'm working with or who worked on the project before me has really messed up the code base in ways I'm going to be dealing with for years? And how hard am I going to have to work to not be that guy?
Finally, the article uses an example of parsing LaTeX with Perl vs. Python, and uses a regular expression. Now, I don't know if LaTeX can, in general, be parsed with regular expressions, but writing a parser aside, it's no surprise that Perl wins when dealing with regular expressions. That's a fine thing, and definitely something Perl's got going for it (or possibly against it, depending on how you feel about regular expressions), but Perl just about grew out of being a regexp DSL -- it's not really a fair fight. Put a full Perl TeX parser up against one written in Prolog, and I bet Prolog would win, but it doesn't mean I'd use Prolog over Perl for most things.