From Figure 6: Weekly activity log at baseline, and at the conclusion of the intervention
Hours active: 41, increased to 46
Hours asleep: 55, decreased to 46. (average # of hours of sleep per day 7.9 hours, declined to 6.6 hours)
Hours sedentary (eg lying on sofa): 72, increased to 76
The paper noted:
It can also be noted that she is sleeping less and in a slightly more disjointed manner at the end of the study.
Yeah. And she has learned to 'decrease the impact of gravity'

when carrying out active tasks (eg use a perching stool when hanging out washing). So it sounds as though the active tasks became more sedentary.
Now, I'm probably giving this paper more analysis than a study of one person who may or may not have ME/CFS with an intervention of one hour of exercise a month warrants. But, if I had to make a conclusion, it looks like the person has managed to achieve a small increase in the number of active hours by increasing the amount of time she rests and by making the active hours less active. And this has resulted in sleep that is now less than the accepted optimum length and is more fragmented.
The authors note that this patient was extremely well supported as she lived at home with her parents.
Therefore, she really could recuperate in between the exercise intervention classes and allow supercompensation to occur.
whatever supercompensation may be. It sounds rather like she collapsed in a heap after the monthly 1 hour of exercise. That suggests that if the activity log had been done in the week following the exercise class, the number of active hours would have decreased from the baseline significantly.
Following this case report, the exercise intervention documented in this case report was undertaken with a group of six individuals affected by CFS/ME and POTS over a six-month period at the CRESTA clinic. The overall conclusion from this was that the effect of the intervention was variable, but for the patients where an improvement was seen, the impact was momentous. The patient's background, medical and behavioural stability were significant components in their progress.
Hmm, so really not replicable then. But, if it didn't work, it wasn't the program's fault.