I wasn't making an argument to that effect.
I think we can talk about purposiveness without metaphysical or anthropocentric baggage, without positing it as a suspension of or exception from materiality, and without requiring the universe to be ever so much more freaky than we ever suspected, man. The language we need for this is the language of cybernetics, which identifies situations within a causal framework that we can meaningfully describe in terms such as "X in order that Y".
The alternatives are fairly unappealing: declare that wherever X happens in order that Y some freaky counter-causal power is at work, or declare that it is never meaningful to describe any situation as one in which X happens in order that Y (or that it is "meaningful" only on the basis of some fundamental misrecognition of what is really going on - I'd rather be eliminativist about "meanings" of this sort, given the choice).
So: X happens "in order that" Y whenever X (a process) is governed by a representation (Y') of a goal (Y) such that Y' is not a complete set of causal antecedents of Y (in other words, it takes something more than just Y' existing to bring Y about) and X maps Y' to Y with sufficient fidelity that, ceteris paribus, variations in Y' correlate with some statistical regularity with variations in Y. That is, roughly, the sense in which we say that a particular allele at a gene site is "the gene for" a particular characteristic expressed in an organism's phenotype. We don't mean that the gene directly determines the characteristic, but that it strongly correlates with the characteristic thanks to its "steering" function in the causal process that actually produces the characteristic.
Is this an anthropocentric projection of some linguistically-generated human illusion of purposefulness onto a fundamentally purposeless process? No, because what we are defining here as purposiveness is not a magical quality that we attribute to things that kind of sort of look purposeful to us thanks to our wonderful powers of ontological delusion. We're coming at this from the other direction: giving an account of purposiveness as a contingent property of (conceivably entirely deterministic) physical systems, and using that account as a starting point for describing what happens when human beings entertain purposeful notions and act on them (or come to believe, somewhat after the fact, that this is what they have been doing). Obviously the human case is about a gazillion times more mediated. But there is still a statistical correlation between my thinking "mmm...donuts!" and my eating donuts, which involves language as only one in a tangle of exceedingly convoluted causal circuits mapping the representation to the outcome.
