This is absolutely a real and important effect, but we should bear in mind that this poll isn't the thing proving it so it's kind of a bad headline. In particular, the headline suggests that this is a new and tentative finding rather than something that's been known for ages, and that it's possible to disprove the effect by knocking down this survey. Intergroup contact theory actually goes back to the 50s and AFAICT is incredibly well-established.

To prove the effect exists with a survey like this, you would need to carefully disentangle the people who "don't know any trans people" because they don't know any trans people from the people who "don't know any trans people" because all the trans people in their life are terrified of coming out to them. Conversely, you would need to carefully disentangle the people who "know a trans person" because they know someone who's out to the world from the people who "know a trans person" because they know someone who's out to only a very few people who they already had good reason to believe would be supportive. There are ways of doing this for people who are better at statistics and experiment design than me, and as I understand it there are studies which do it carefully and do prove the effect, but this isn't one of them and doesn't try to be. (And why should it try to be, when the effect's existence has already been established and studied separately, and when the raw data on a large current sample is useful without reinventing the wheel?)