This almost weakens, but not quite: because those people who wanted to be in the chocolate group were nevertheless randomly assigned, the experiment remains intact.

This strengthens the author’s argument by affirming the author's assumption about representativeness. Another study with the same result makes it that much more likely that the findings are true for all people.

We have no reason to think that the difference between solid and liquid foods is relevant to how people perceive taste, or whether tastes can interfere with each other. As far as we know, this isn't relevant.

In other words, the chocolate group might simply have no ability to taste the differences between coffee samples. This denies the author's assumption about relevant differences, and weakens by presenting a new factor that better explains the difference between the groups.

Regardless of the strength of the differences, these people all noted differences, whereas people in the chocolate group didn’t note any differences. This doesn't change the result or add new relevant information, so it's not useful.