Let's look at the numbers.
98% of incarcerated sex offenders are male. They make up about 20% of the male prison population. That 20% outnumbers the entire female prison population, most of whom are in prison for financial/property/drugs crimes rather than crimes against the person.
Fair Play for Women did an analysis of trans identifying prisoners in the male estate using FOI requests and official documents (their numbers were later confirmed by the BBC fact check people and by the government). There were, at the time of the analysis, about 120 prisoners in the male estate said they were trans (I've chosen that wording carefully for reasons which will rapidly become apparent). Of that 120, 48% were convicted sex offenders.
This means that prisoners in the male estate who say they're trans are more likely to be sex offenders than randomly chosen male prisoners (nearly 1 in 2, rather than 1 in 5).
Now at this point either we accept that people are always exactly what they say they are and these 120 prisoners really are trans (acceptance without exception) or we entertain the possibility that some or even most of them are lying.
Let's look at the two horns of this dilemma.
Now, if they're all telling the truth, then i would argue this is a pretty strong argument against treating TW prisoners as women and moving them to the female estate - statistically, they are hugely likely to pose a risk to women prisoners (who are typically highly vulnerable, with a much higher likelihood of having been victims of child sexual abuse, sexual abuse and domestic violence than women in the wider population).
Alternatively it could be that many of them are not trans at all but are in fact lying to gain access to women's prison either to gain access to a literally captive pool of potential victims or because they think women's prisons will provide an easier way of doing time. Incidentally both the association of prison governors and the professional body of criminal psychologists who gave evidence to the parliamentary committee chaired by Maria Miller warned of this danger: that predatory men might pretend to be trans to get transferred to women's prison. (It's all there in Hansard but for some reason was left out of the final report). In this case, self ID looks like a spectacularly bad idea.
In short, either you accept that TW in prison pose a massively higher risk of committing sex offences than women, or you accept than sex offenders frequently lie about pretty much everything including being trans.
In neither case does accepting self ID look like a good idea when viewed from the perspective of the human rights of women prisoners.