The quote that bothers me somewhat is this one:
The data here is on the very edge of reality, built on too many assumptions.
Data is data. Assumptions are the stuff of models and theories. Don't mix the two.
Data is nothing if you do not have any way to interpret it. Models and theories provide the context for interpreting the data.
It is like saying "bits are bits, assumptions are the stuff of encoding and decoding". Problem is, without any assumption to decode your bits, it would be as useful as any random noise. The fact that we can have a conversation here is because I (or rather, my browser) made the assumption that the bits are encoded with a certain pattern, and so did you.
Without any assumptions, models, or theories, the signals we received from Hubble would be no different from random noise.
Without the assumption that the photons came from a distant galaxy, we cannot form the image we can see.
Without the assumption of what they saw were the result of the collision of two galaxies, it would just be a bunch of stars in a strange shape.
Without the assumption of the current model of our universe, we cannot guess what would be the most probably original form of the two galaxies.
Without the assumption of the Theory of Gravity, no one can make sense of what could have happened when two galaxies collide, and thus compare with this observation.
Without the assumption of the model of gases and stars, we cannot reach the conclusion that gases should interact and slow down, while stars would not.
The problem is, with our currently best assumptions, models and theories, those that are able to explain most of our observable universe, we found that it would require the present of some undetectable matter in all the galaxies to make everything consistent -- hence "dark matter".
Yeah, you can claim that is too many levels of assumptions. Feel free to build up your own that could consistently match all the known data even better than the one commonly used.