Was just looking into the crossover January weeks and found it is not consistant (or maybe the method eludes me):
So a week that ended Jan 4th was part of the previous year, then the next year a week ending Jan 3rd was part of the new year
*Edit: Just checked and it isn't actually counted in Dec 2014 or Jan 15... but instead added into Dec 2015 for some reason.
It is very interesting what you point out. Obviously with all of these numbers and having to revise them will cause small Errors. It kind of seems like Months and years are calculated independently, or even manually. I got curious with your discovery and decided to do a little comparison between VGC months and using the method of giving the week to the month that has more days. It is obvious that I found many diferences, but one interesting thing is that some ocasions something that was added to one month ended up being distributed across at least two different months. This makes me wonder if it was manually edited.
It might be too much to share the whole chart here, but I highlighted some occasions where something was given to one month but then two months got that value distributed between them. Like PS3 months 3,4, and 5. In General these little details are minor numbers that probably got manually assigned. But something that is important is that all of these green and red numbers show that the system automatically just gives the week to the month it lands on at the end date. This might be something that @PAOerfulone might find interesting.