People in general often misstate their preferences, or their behavior fails to match it. According to research summarized on my blog, both men and women do this, and women on average just do it more.

From Urbaniak, G. C., & Kilmann, P. R. (2006). Niceness and dating success: A further test of the nice guy stereotype. Sex Roles, 55, 209-224. (emphasis mine):

Weiderman and Dubois (1998) used behavioral measures to assess women’s preferences for a mate and found a discrepancy between self-perceptions and behavior, particularly among women. For both men and women, the physical attractiveness manipulation was the most important factor in predicting ratings of desirability. Men accurately indicated that the physical attractiveness of the targets was the most important characteristic that influenced their desirability ratings, whereas women inaccurately indicated that desired level of relationship commitment was their most important factor, when, in fact, it was one of the least important factors behaviorally. Sprecher (1989) found similar results, in that women inaccurately assessed the role of physical attractiveness in their own ratings of a target man. The women in Sprecher ’s study reported that expressiveness was the most important factor in their choice, although it was the least important factor behaviorally. Physical attractiveness was the most important factor that actually influenced their ratings. The results of these two studies suggest that women’s self-reported preferences may not match their actual choices. Because it is still considered shallow and inappropriate for women to say that physical attractiveness is very important in their choices, those women may have engaged in impression management.

From the Sprecher article:

In a classic paper in the area of social cognition, Nisbett and Wilson (1977) argued that people often do not know what stimulus creates a particular response, and in such cases use “implicit causal theories” provided by the culture to explain a response. [...] The same lack of awareness argument could be applied to this area of determinants of initial attraction. Identifying what characteristics are desired in a partner and why attraction is or is not experienced toward a specific person involves higher order cognitive processes that people may be incapable of successfully monitoring. Instead, men and women may rely on implicit causal theories or social belief systems to determine what they report to be attractive in someone. Consistent with this, Duck and Sants (1983) have argued that personal relationships researchers attribute more self-awareness to participants in relationships than they actually have.