My question is whether propensities are one of the building blocks of purposes. Purposes are pretty foundational in action -- we do things because there's something we're trying to achieve or make happen. Whatever it is that we're trying to achieve is the purpose. Clearly, though, our actions are at least somewhat caused: we're biological creatures, we have physical bodies, there's all kinds of complicated electrochemical activity going on all the time. That's all causal material -- there's no purposes involved in neurons firing in particular orders. There can, however, be propensities: a certain disjunctive set of electrochemical structures (for simplicity, I'll presume in the brain) can be held to constitute a propensity, something that increases the causal likelihood that certain events (movements) will occur.
So, we have propensities operating at some minor abstraction from a physical level, and purposes operating at much higher, much more abstract level. What's the connection between them? They're clearly not the same thing: propensities are what propels one into certain movements, purposes are what one achieve through certain actions. Even if we equate action with movement (usually a bad move), the former are still backward-looking while the latter are forward-looking. This characteristic difference in perspectives demonstrates that the two are not actually the same. Presuming that we don't want to eliminate one in favour of the other, there needs to be a way to link the two together.
My sense is that the connection has to be agency. It is only agents who can get purposes out of physical propensities. And agents do this because of the power they have, namely agency itself. Agency itself I have already argued is a form of causal influence: agents, through exerting their will (the power characteristic of agency), can change the course their causal histories are in the process of determining. Propensities are the causal history of the agent: they capture what it is the agent is likely to do, assuming no interference by the will. Purposes are what the agent's will is directed towards: the agent is attempting to turn the tide of causation, so to speak, towards a particular goal. So, to get from propensities to purposes, we need an agent's will (agency itself) that encourages certain propensities to be actualized, and inhibits certain others.
This implies that there is a non-natural element involved in action: namely, the will. I know of no way to analyze the will naturally that doesn't collapse it into some complex set of propensities, which is one of the eliminativist moves I was trying to avoid a few paragraphs ago. However, this is not non-naturalism or supernaturalism all the way down: propensities are still analyzed naturally.
From this, it also follows that we can offer a teleological model of action-explanation -- that is, explain actions by citing purposes -- that is non-mysterious (constituted by natural purposes and the non-natural will) and yet still has the full explanatory potential characteristic of teleology.
So, in short, we don't have to give up anything in order to fully understand actions: we don't have to give up our sense of ourselves as largely non-natural beings in a world of purposes, nor do we have to give up our sense of ourselves as largely natural beings in a world of causes (propensities). Furthermore, assuming that the teleological model is a better model of action-explanation than competitors, we have as much predictive and explanatory power as we could ask for. And, finally, actions end up being a special case of events. Events occur when the propensities are not interfered with. Actions occur when the will exerts itself (possibly only successfully) upon the propensities.