The point of a biomarker is that we think/hope/hypothesize that we can use it as a surrogate for the thing we're actually interested in, so that we have something measurable on shorter timescales. Unfortunately, this assumption is often incorrect, and we won't actually know for years.
How do we often test for diabetes.. cancers.. many many clinical conditions? Biomarkers. So they are not practically unreliable in general.
What do you know about the particular biomarker employed in the study to say that it is unreliable ?
Long term degenerative conditions are some of the most difficult problems for such validation, due to time, cost, and the complexity of risk in real populations.
I don't know anything about how well the current study actually links to mitochondrial biogenesis, but I know that we don't have satisfactory interventional biomarkers for a process as complex as aging. By all means continue researching this, but I think it's eminently sensible to be skeptical at this point.