Alchain-Allen being true in most or every model of economics (as the Church-Turing thesis [1] is true in most models of computation), or being always true in reality, is a much more debatable question.
Real demand curves can be mighty pathological, for example, you can have D'(P) > 0 [2] if the good in question is something people buy to show off their wealth, and so actually like it to be more expensive.
[1] http://en.wikipedia.org/wiki/Church_Turing_Thesis
[2] Economists like to put demand on the horizontal axis and price on the vertical axis. That convention is insane, so I ignore it when I can.
If the independent variable is price, then your thought process will be: "If the price is $100, then there will be 1000 buyers. If the price is $150, then there will be 700 customers." It's easy to think about things that way.
If the independent variable is demand, then your thought process will be: "If there are 1000 people buying the product, it must cost $100. If there are 700 buyers, then it must cost $150." It seems counterintuitive to think in this way.
[1] Except in economics, of course!
[1] http://en.wikipedia.org/wiki/Theorem