First off, I do want to say that I personally dislike the idea of tiering megas separately, and that I agree with bughouse that this very much seems like a case of bias throwing logic out in order to get a desired short term result without care for the long term. Maybe there won't be any long term problems, but the fact is that our tiering system is made to place Pokemon in the spots they have earned, and not where certain users think they should be, and a lot of what I have seen here is people arguing for this because they think it would be better if Pokemon were in different places than they are now. In other words, while I oppose this, I don't necessarily think it is any worse than what we have. Rather, I just think there has not been any real convincing reason given to do so, using the logic that we use for our tiering system.
That said, if we do go with tiering them separately, I think that simply trying to use the existing cutoffs as we have them for megas is foolish and ignoring the reality of the situation. I know a lot of people like thinking of Megas as their own Pokemon, and in most ways that is a fine way of thinking of it. But, it is simply not the truth, and when it comes to measuring the usage, that can have a major effect. The fact is that a mega is just a specific item on a set, and just like any set, a set with a mega stone faces competition, not only from other Pokemon, but from other sets that the same Pokemon can use. If we tried to apply the logic being used to any other item, the argument would easily fall apart. I mean, clearly, while a Life Orb and a Choice Band set of the same Pokemon function differently, they generally will be used to take on similar Pokemon, and if one item didn't exist, the other would see more usage on that same Pokemon.
This is relevant when it comes to megas due to Pokemon like Gyarados or Alakazam who, as whole Pokemon, have OU level usage, but when you separate out megas, do not. To call either form not OU simply by using our existing statistical measures is foolish, because it ignores the fact that this is never done elsewhere with just a set. It calls all the other traits they share irrelevant, and ignores the fact that if the mega did not exist, a significant portion of the usage could potentially stay with the Pokemon itself, just with different sets.
Now, I am not saying that there is no good way for us to tier Pokemon and their megas separately. In fact, I'm sure there is. All I am saying is that our existing system is made to tier Pokemon, not sets, and yes, no matter how you think of them, Mega Pokemon are mechanically just sets. I'm not a big stats person, but I would just say that if this is the way we ultimately want to go, we should ask those of us who are stats people to see if they can come up with a new statistical measure that better works for tiering the usage of sets, and doesn't cause any arguably OU Pokemon to drop just because of self competition.
EDIT: For clarification, I am not specifically saying that Gyarados or Alakazam should not drop. Just that the current system does not provide good enough justification for such an occurrence.