In my previous blog post Behat - The Right Way I made a statement that I think Behat was a better choice for writing tests even for the frontend. Some good arguments were raised in favor of CasperJS.
I believe my comparison was wrong in the sense it was lacking the key point to Behat's stength for us. It's not really about "Behat vs. Casper". The proper comparison should have been "Behat vs. Casper - With a Drupal backend"
And here's the key difference: With Behat you can interact with Drupal's API even when testing using PhantomJS. That is a lot of testing power!