Bioware seems to have enough pull to be able to make most of its own decisions because they haven't slipped up yet but other developers simply don't have this advantage so publishers make huge demands from the start and effectively cripple most games before they even got a chance to prove themselves to be able to make great decisions on their own.
When I read an interview about "the making of" of games in the past, it nearly always includes a part that says "well we wanted to do this but the publisher told us not to" and more often than not, what they wanted was exactly what the game was missing to be great.

EA seems to be like that with at least most of its studios now though. I even recall one instance where Riccitiello was saying how Mirror's Edge "could never work" but they tried it anyway, and they've since approved a sequel. :)
A particularly funny example I always think of is from CD Projekt:
"Other more general fundamental ideas were for instance that we'd make an edgy game with its distinct character regardless of political correctness and current trends. What's more, we insisted that no matter how much we respect the advice of experienced publishers, at the end of the day we've got the final word on anything regarding creativity. We didn't want the vision of the game to be in any way distorted or damped. An odd example of that occured during a conversation with a publisher who said that on account of their market research players overall want their protagonist to be an elfish woman and that if we had changed The Witcher accordingly they would have considered negotiating a contract :). We have been defending our vision all the time. Fortunately Atari perfectly understood our concept and even added a couple of nice ideas which improved the game. In contrast to other publishers they were happy we had such a solid and clear vision."
And what do you know, they sold 1.5m. On PC. So what the hell do publishers know. :P
And yeah. Money doesn't mix with games, or any creative expression. There will always be that conflict so long as they co-exist.

I think it mentioned on one of the GOG editorials here, that for example, KingPin was originally going to be a much much bigger game, involving full scale gang warfare, something akin to a RTS game, but at the end everything was slimmed down due to Interplay's requests, and ended up the final product we all know. I think it was mostly because of time and resources problems.

I completed Kingpin last weekend, oddly enough. It's funny but I sensed there was supposed to be more to the game - you could say "yes", "no" and "special" but you never needed the "no" since there weren't actually any real convos. Also, your team mate AI seemed to be heavily underused with only three occasions of them actually being offered to you.
In general, you need to balance the developer's wild ambitions with common sense which publishers are *supposed* to bring i.e. don't overdo things that would lead the game to take too long to make but what most publishers seem to do, is try to cut the risk which, ironically, ends up making the game no longer unique which harms sales.
Did you know that Quake was originally supposed to be more like Half Life? Romero actually left because he had big plans while the others just wanted to throw in Doom weapons, basic level design, and get it done with. In this case, it wasn't the publisher's fault but it goes to show that just taking the "easy & quick" route is rarely good: Quake SUCKED as a single player game.