Prior to the intervention, the two groups were randomly selected. However, the goal of the study was to estimate the magnitude and direction of social approval bias in reported fruit and vegetable consumption obtained from a short food frequency questionnaire and 24-hour recall. Thus, an effect was expected following the intervention.

Thanks for the editing!

Regarding #2: In the study discussion, the authors question whether the difference in behavior could reflect an actual change in behavior (your first explanation) or an intention to change (your second explanation). They suggest that the minimal nature of the intervention and the large size of the difference make those explanations unlikely.

1. Investigators send information promoting fruit and vegetable consumption.

2. After receiving this information, people respond in the short term by eating more fruit and more vegetables; or, they become more aware of their fruit and vegetable consumption and more likely to remember it; or, they deliberately exaggerate their fruit and vegetable consumption in order to impress strangers.

(The three explanations are all plausible; I present them in what seems to me to be the most likely order of probability.)

3. “Patients lie! They lie! They’re all lying liars!”

Now, this:
“Since the two groups were randomly selected, fruit and vegetable consumption should have been the same in each group.”
clearly shows the mistake of interpretation here, since the investigators performed an intervention on one group before sending their questionnaire. There’s no reason to think their intervention would have been without effect, so no, the consumption most likely should not have been the same in each group.

So the study authors can’t be held blameless. But what’s your excuse?