Because regardless of how you feel about sexual orientations, it's not common that you run into 9-12 people who all happen to be bisexual. Maybe "unrealistic" isn't the right word, but it's certainly not typical of the people you would meet in real life. You may have 9 or 12 awesomely badass friends, but I highly doubt that even half of them are the same sexual orientation. One of the amazing aspects about people in general, is that two individuals can be vastly different from one another. Like Samara said, "If there are three humans in a room, there will be six opinions." The idea that the sexual orientation of that many people would be exactly the same strikes me as unlikely, even in the fantasy setting of Dragon Age.
I do not like the insistence that the party make up is incongruent with whatever people may happen to think is an acceptable distribution of sexual orientation and therefore that takes away from the game. Forgive my bias, but doubly so because in my experience there'd be less of a concern if all the companions were hetereosexual.
Lets assume there's a straight up 1/3 1/3 1/3 split of heterosexual, homosexual, and bisexual. With 4 romanceable companions, that means that the odds of there being 4 love interests that are all of the same orientation is about 1%. Which may seem like a small number... but it doesn't strike me as being frighteningly shocking. Unlikely? Sure. So DA2 was unlikely then. I'm still not really seeing the problem. Note that this is the odds of them all being heterosexual and homosexual as well (does the number seem a little too small for heterosexuals for some people? I wouldn't be surprised if it does).
Now, lets look at the probability of there being 2/1/1 (in whatever order).
Probability that the orientation of the first is one of the three is 1.
Probability that the orientation of the second one is NOT the same as the first is 2/3
Probability that the orientation of the third one is NOT either of the first two is 1/3
Probability of the 4th is irrelevant, so it's 1.
So the probability of each of the permutations is 21.8%. Which isn't really super high individually.
Now, I know what you're thinking! "But Allan, we can take different permutations and escalate that number." And you're right. So the odds of a nebulous defined 2/1/1 is only really useful if we are procedurally creating randomized orientations (which we're not).
So if we were to imagine a Dragon Age 2 that is, say:
2 people are bisexual
1 person is homosexual
1 person is heterosexual
There's only a 21.8% chance of that particular grouping actually occurring. The problem is that people don't see this. They see it as being equivalent to:
1 person is bisexual.
2 people are homosexual
1 person is heterosexual
and
1/1/2
But they're not. In their mind their seeing the 65.3% and go "that's the way it should be based on probability." But you're not getting nebulous 2/1/1. You're getting a specific one, and with that comes a specific probability which is 21.8%. Suddenly that specific match up you're given, while still more likely, is hardly "definitive."
So 0/0/4 (which is appealing because it's seen as universally fair) is about 1%. We could have gone 2/2/0. lets look at that.
Odds that first person is not bisexual is 2/3
Odds that second person is not the same orientation as first and not bisexual: 1/3
Odds that third person is not bisexual: 2/3
Odds that fourth person is not the same orientation as 3rd and not bisexual: 1/3
This is about 4.8% for the specific grouping that you see to actually exist from 4 mandated characters that you know will exist and could be potentially 1 of 3 equally distributed distributions.
(This last one I may have effed up on.... I may have gotten some conditional probabilities wrong in my head, but for the most part it's 50% of 2/3s which is 1/3 and it makes sense at 1:15 AM....)