Don't confuse study design and author behaviour with the questions being asked.
If you look at the actual reasons for retracting his view, it was based on further evidence failing to reproduce the work.
The initial study was 'good' - but for the sample size and protocol.
It did not use self-reported outcome measures, but tried to measure actual concrete things.
It measured hormone levels pre and post, as well as actual behaviour on a task.
At the least, if true, some aspect of the 'power pose' affected hormones.
The brain affects the body in real concrete ways. People with lower relative incomes, or who are stressed die earlier.
I have not read the paper, but if in fact they as the coverage mentioned did a test, then as nothing was showing, added more people and re-did it, that is enormously statistically risky, which makes the protocol bad.
The actual fundamental question was good in that looking at all the evidence obtained by looking at this, we can say it probably doesn't happen, when some of the literature before the study suggested it might.
The study design was bad.
The question asked in the PACE trial was a good one 'Does APT/CBT/GET/SMC help CFS patients more'.
The study design was not great - it did not have enough non self-reported outcome measures, but not terrible.
The authors behaviour in modifying the protocol and in sending all participants 'CBT is awesome' leaflet partway through was terrible
PACE would have been enormously positive - if the authors at the time had stuck to their protocol and reported 'there was a small, but statistically insignificant benefit for CBT and GET compared to SMC. There was a small but statistically insignifcant worsening for those on the APT arm. The initial hypothesis of this study was not proved, and further research is needed before recommending any of these therapies.'
Imagine how many thousands (tens of thousands?) of patients would not have had CBT to no avail.