I don’t think you understand that different kinds of data answer different kinds of questions.
Individual descriptions are entirely appropriate evidence for understanding subjective experience. If you want to know what a particular phenomenon feels like to people who experience it- first-person accounts are directly relevant. When similar descriptions recur across many people and are also reflected in clinical literature, they can help characterise a recognised psychological or medical phenomenon.
What individual accounts cannot do is establish prevalence or comparative population rates.
So I can reasonably say, “this is how trans people consistently describe the experience of being trans,” because personal accounts are evidence about experience.
I cannot say, however,m, “30% of trans people experience severe gender dysphoria” on the basis of collecting individual testimonies from people who describe severe dysphoria. To make that claim, I would need data capable of estimating the proportion in the wider trans population.
Similarly , personal testimonies can establish that trans people sometimes offend and provide evidence of how that is experienced at a personal level . It cannot establish the rate of offending among trans people as a population, or show that this rate is higher than in another population, unless you have an appropriate denominator and representative or otherwise valid population-level data.
That isn’t “picking and choosing” evidence. It is basic methodology: the evidence has to be capable of answering the question being asked.