I haven't trawled through the papers in detail, but I think the short answer is that in the Andromeda measurements, a variety of methods have converged, whereas in Triangulum, they don't agree as well. For example, if you look at standard methods like observing Cepheid variables or using the tip of the RGB, the errors are both quoted as 40 kpc or 25 kpc, respectively. But in Andromeda, the corresponding distances are 770 and 780 kpc, whereas in Triangulum they're 850 and 790 kpc.
Note that the Wikipedia page uses the eclipsing binary observation too. That gives the upper limit (940 kpc) so who knows if there's something about the stars' light that messes one about. Also, the Wiki page quotes a range instead of combining the averages properly. I'm sure a proper statistical combination of the measurements would give about 860±46 kpc. Still not as precise as for Andromeda, but better than the 105 kpc implied by the range.