It seems likely to me that they asked that because they felt respondents would be most likely to recall what they’d eaten most recently, and less likely to be able to do so outside of that window.

Not sure what the real problems are extrapolating that that one day is a typical day given that routinely you extrapolate from a small sample, if that sample is representative, to much larger populations with pretty acceptable margins for error.

I guess the question is, was this question asked at a time or in a manner that would lead us to know that the period being asked about was in some way atypical for the population being asked.