I thumbed you up because you were technically correct about the fact that just because positive judgements drop doesn't mean there's a bias.

However, there is some extra data in this economist article on the same study to support the idea that there weren't factors in the arrangements of parole candidates that would account for such a drop:

To be sure, mealtimes were not the only thing that predicted the outcome of the rulings. Offenders who appeared prone to recidivism (in this case those with previous convictions) were more likely to be turned down, as were those who were not in a rehabilitation programme. Happily, neither the sex nor the ethnicity of the prisoners seemed to matter to the judges. Nor did the length of time the offenders had already spent in prison, nor even the severity of their crimes (as assessed by a separate panel of legal experts). But after controlling for recidivism and rehabilitation programmes, the meal-related pattern remained.