Just read the paper. In the 2149 participants, detection in fasting plasma was associated with a hazard ratio of 1.57

It was also shown to enhance clotting at the levels detected in fasting plasma (in 10 people) which provides a potential mechanism to explain the effects observed in the larger group.

As for how legit it is... The number of people is significant. My only complaint is they didn't measure xylitol directly, rather they tested for polyol or something.