Stuff like this tends to come up occasionally that ultimately leads nowhere. Usually, if there was something substantially off with the current tracking, it would already be detected by multiple people. Remember the inaccuracies of laser sensors? That was spotted quick (unless I'm just too old and it felt like it was fast ). The Razer CPI stuff with the Viper was also discovered pretty quick, and even people who didn't know in particular what it was would say it felt pretty fast compared to other mice at the same CPI - says standard deviations. There isn't anything widespread or noteworthy about the input being filtered.
I don't think the suggested edited XMLs do anything in particular besides further manipulate tracking, giving the poster a belief in something positive. It's just more feels stuff. There was already posts showing that the Logitech software tracking output are pretty much the same with software uninstalled. Also, they tend to be the benchmark for tracking anyways.
If they come up with some output comparisons on tracking, I'll be interdasted. Until then, I'm inclined not to believe.
That phenomenon is nothing new. I've noticed it happening on every brand of mouse and their own respective software packages for years. It's not a Logitech only issue (could be a Pixart-specific issue though?). Installing the software makes it so you'll never get it to feel how it did straight from the factory.
I had a similiar issue when I turned on surface detection in the Asus gladius 1, it altered the mouse feeling in some way. I was able at any time to turn it off though. Maybe thats build into logitech g hub?
A lot of people have the habit of looking through configs and then changing parameters without knowing what exactly they do. They assume things based on the name of the parameter, even if its misleading or that parameter isn't even active because of some parent parameter controlling it.