* we recognize that the optimal long-term strategy can differ greatly from the optimal one-shot analysis, and
* we have preferences about some anticipated world-states rather than just anticipated mind-states.

In response to this I say there is nothing about subjectivist satisfaction which prevents taking these (or anything else) into consideration. Further, I do not mean this in a utility-function sense, but rather 'actual wants derived from valuation forecasts which result in intentionality'.