If severe mental health problems is considered a good reason to deny blockers, that strikes me as a fairly significant confounding variable even when sample sizes are ample.
Yes, WPATH guidelines state that mental health problems should be controlled before commencing hormonal interventions (although there is evidence that not all clinics, especially in the US, do this consistently). No study has used random assignment to medical versus psychotherapy intervention. If there is a comparison group it's usually a group that weren't given blockers/hormones even if they were seeking them, and there is often a lack of information about the reason.
The selective attrition is particularly striking in the Tordoff study though.