My response to this kind of pretense theory is usually along the following lines. At the moment, no actual scientist has used the Santa Claus story to compute the Lamb shift, analyse quark confinement or perturbations in binary star systems and so on. So, I would be very interested to know why not!
The basic claims of such pretense theories are in some sense connected to the meanings of mathematical sentences; but they are neither abstract metaphysical claims (e.g., "mathematical facts are necessities which trivially supervene on all contingencies"), or rather abstract epistemological claims (e.g., normative claims, concerning rationality, "oughts", and so on). What is interesting about such claims is that they seem to have some empirical content at the level of cognitive psychology, and therefore may be subjected to empirical investigation.
In an interesting forthcoming article, "Pretense, Mathematics, and Cognitive Neuroscience" (BJPS 2013), Jonathan Tallant argues:
A pretense theory of a given discourse is a theory that claims that we do not believe or assert the propositions expressed by the sentences we token (speak, write, and so on) when taking part in that discourse. Instead, according to pretense theory, we are speaking from within a pretense. According to pretense theories of mathematics, we engage with mathematics as we do a pretense. We do not use mathematical language to make claims that express propositions and, thus, we do not use mathematical discourse to make claims that are either true or false. In this paper I make use of recent findings from cognitive neuroscience and developmental science to suggest that pretense theories of mathematics fail.