That being said - I've not seen much consistency across implementations. It generally breaks down to some sort of product backlog which is worked on in small chunks on some cycle (months, weeks, two-weeks whatever...). This is a pretty basic change to the normal waterfall cycle and easy to implement, and works well for many teams.
The rest of the Agile/Scrum/pick your particular method is left to the project/product/SE manager and you really have to find what works for your team, and I'm not sure makes that big of a difference in the end.
But all real-world instantiations of methodologies are, in some sense, "amalgamations". For one thing, none of them are complete. As both the article and Raganwald point out, Scrum is far from a complete description of what you need to do to organize a successful project, just as using Lisp and Emacs aren't sufficient conditions for producing great software.
My biggest problem with that is that I think a lot of these companies are having their people learn Scrum because they think it'll be their silver bullet, as if taking a bunch of people and having them start doing Scrum will suddenly turn all of their failed or ultradelayed projects into victories.
Scrum worked very well for us at MTV Networks with the project I was working on. Due to us having a very experienced Scrum Master and the Project Owner being very active and understanding how to prioritize his requirements.