Some things to keep in mind when trying to compare statistics on this:
A. If we wish to compare rates of a particular type of violence between victims belonging to different demographic groups AND if we wish to assign any observed difference in the level of that violence TO the memberships in those different demographic groups, then, ideally, we would compare people who are otherwise the same except for that membership difference.
This is very hard to do in practice, but trying to control for other things which affect the likelihood of becoming a victim of violence is a way to get closer to that. So, for instance, one reason that men have a higher rate of victimisation, on average, than women, is because men are more likely not to try to mitigate the risk.
In other words, men are more likely to go out at night alone, get into fights with other men when drunk, engage more in the kinds of criminal activities which increase the risk of violence, and perhaps also act in a more aggressive manner towards other men (than women towards men) which is more likely to cause a violent retaliation.
When we apply this to comparing trans women to women who are not trans, the same problem applies: There are other average differences between the groups which should be controlled for before drawing conclusions from the findings.
B. Getting data on people becoming victims of violence is difficult, in general. The two main ways we do that are either by asking people about their experiences or by using information that has come to the attention of the police.
Both types have problems. Not all cases are reported to the police, and we know that most rapes are not, for instance. But self-evaluations of victimisation can also be problematic because there is no other evidence and because it is possible that different demographic groups might define what they see as violent behaviour differently.
When comparing trans women to women who are not trans the problem may be no different, but much of the data on transgender victims has not come from the same sources as the general data (often this research is done by activist organisations such as Stonewall or Glaad or the Trevor Project), and this makes it possible that the questions used are different from what general surveys would use.
c. There is a disconnect between homicide statistics and other statistics about violence. Data from the US and the UK, at least, do not show transgender individuals facing higher rates of homicide than the populations in general, but data from some sources suggest that they do face higher rates of non-lethal violence. I wish to understand this discrepancy, but I have not seen any good explanation for it.