I am using Alexa data in my dissertation and I think it can be reliable for some things.  It is just an issue of understanding how Alexa gets its data, and then seeing if the results are similar on compete and Quantcast.  In some cases, you don't really have much of a choice but to use Alexa.  For Australian and New Zealand website rankings, I can't think of another site that provides publicly available free data.  I've used it in a blog post at http://ozziesport.com/2011/02/new-zealand-sport-web-traffic-in-response-to-the-christchurch-earthquake/ and for general event tracking, it can be useful to see how events impacted a site's traffic amongst the group that uses it.

I just wouldn't use Alexa data in isolation involving other data.  I used it in http://ozziesport.com/2011/01/derryn-hinch-journalist-traffic-versus-wikipedia-traffic-in-response-to-st-kilda-controversy/ (which has broken images) in a bit that I considered using for my dissertation chapter to try to guess at the amount of web traffic that a journalist got.  So yeah, Alexa data can be used and in some cases should be used but against a backdrop of other information.