I think your thought process brings up a few different aspects of evaluating ethical philosophies, and disentangling them would be very helpful.

First, I certainly agree that there are probably people out there that reach utilitarianism through a process of motivated cognition - they want to be in control, and the reason they use (perhaps even to themselves) to make that sound better is that it would be for the good of everyone. However, I also think that there are many other people out there who grew up believing that good is what we should strive for and that the way to do that is to aim for the greatest benefit of the greatest number of people. These types of people might then reach for utilitarianism not to justify actions they wanted to perform already, but rather as what they perceive as the closest complete ethical system to their previous objectives.

While the former group of people merely use utilitarianism as an excuse (even if they believe they believe in it), it is actually the latter group whose reasoning I am generally more concerned about. Whereas the dictator types will do what they planned to do anyway, the forces of good types are vulnerable to taking utilitarianism too seriously, and doing such things as, for example, thinking that maybe it's okay to sacrifice one human life if it will save one million ants (without considering ecosystem impacts), which I do not think is a thought that would have ever arisen from their core belief system. Which is not to say that all utilitarians would agree with that trade-off, but I have seen some who seem like they would, and that is just one minor example of the many problems I have with the idea.

The other point I wanted to bring up is that utilitarianism is really a system for general thinking, even if one likes it, not for immediate real-world implementation. Indeed, it is unimplementable, in the mechanism design sense. So the only way (that I can think of) that you could put it into practice in the real world is to have a strong AI (or equivalent) build detailed models of everyone (possibly involving brain scans) and implement a solution based on those (as otherwise, any implementation would suffer from participants refusing to tell the truth about their utility functions). So, the question of "how would contentious decisions be made?" is fairly unanswerable, except through accepting some deviation from utilitarianism.

I hope that helps crystallize your thoughts a little bit.