People here didn't understand the Regression Theorem abstractly. The point is to generalize it more, not less.
Try to use praxeology, not other tools, to understand economics.
The praxeological analysis never concern itself with the reasons of actions, only with the effects of actions.
BTW praxeology is the tool used by Austrian economists to analyze the market behavior.

There is a simple explanation of why bitcoins have value and Mises's Regression Theorem is right.

http://wiki.mises.org/wiki/Regression_theorem

"People today expect money to have a certain purchasing power tomorrow, because of their memory of its purchasing power yesterday. We then push the problem back one step. People yesterday anticipated today's purchasing power, because they remembered that money could be exchanged for other goods and services two days ago. And so on."

The "two pizza for 10K bitcoins" is the first recorded exchange of something with a hard value for bitcoin.

The reason the guy gave two pizzas for 10k bitcoins is unimportant. What matter is the act of buying "two pizzas for 10k bitcoins". 

The day after, people (the pizza guys and the bitcoin guy and everyone that knew of the transaction) have a price point to remember.
It doesn't matter if it a good point, a bad point, a neutral point or a random point.
The day after people would remember a price point and they would decide if they want pay less or more to obtain bitcoins.

The same was true for any other good or service exchange for bitcoin. The reason people do the exchange is immaterial. But for every exchange there is a price point to be remembered.
When they started to be enough, these price points started to converge on a smaller band (like the theory would expect) because people try to maximize their expected profits.