The article itself, as Gerry already mentioned, doesn't even claim there is no benefit. It said there was no difference in test scores, but consider the study population. They didn't even start with comparing apples to apples. They started by comparing similarly sized fruit and assuming they were the same thing. Even still, they did find benefits in some scenarios.
Overall it's poor methodology, not very convincing methodology. If they wanted to compare apples to apples, then the study population should be kids with positive diagnoses. To evaluate the efficacy of the medication, they should have used repeated measures, where the unit is one kid, rather than a whole population. This is basic study design that the authors should have a firm grasp of by now, which makes me wonder about their competencies, as in field, not whether or not they're incompetent.
A paired t-test is far more powerful statistically. Individual differences will be factored out of the error term in the analysis. And because it's more powerful, they could have used fewer subjects. That would have left them with more resources which could have been used in any number of fashions, extra tests, better quality testing, etc.