You are allowed to write scientific speculation that counters “currently established fact” – just give us a reason why that makes sense in your universe. (For some universes it can be highly whimsical, for others you’ll need serious handwavium.)
Read as: "Be consistent with your own aims."
Philip K. Dick, god bless his crazy soul, once gave a lecture with the title "How to Build a Universe That Doesn't Fall Apart Two Days Later." In typical Dick fashion, it meanders a bit and takes a rather major detour into the very murky waters of his Exegesis, from which few (Dick included) return intact. But he makes a point I rather like:
I will reveal a secret to you: I like to build universes which do fall apart. I like to see them come unglued, and I like to see how the characters in the novels cope with this problem. I have a secret love of chaos. There should be more of it. Do not believe — and I am dead serious when I say this — do not assume that order and stability are always good, in a society or in a universe. The old, the ossified, must always give way to new life and the birth of new things.
In other words, when things fall apart in a Dick novel, it's not because he isn't paying attention. It's because he's using that as a way to tell the story he really wants to tell. The first few times I ran into this, it did feel much like he was improvising his way into (and out of) situations that didn't seem to have any clear throughline. His lesser books do have that flavor about them, but his best ones follow through on their promise. They are consistent in the way they explore being inconsistent. But not everyone is Philip K. Dick, nor should they be, and so most of the time a story has to be built to be consistent on its own terms or it simply doesn't work.
With SF&F, it isn't too difficult to be internally consistent even at the expense of established fact. It takes an attention to detail, but it also takes the courage of one's convictions — the nerve to believe that your idea, once extended to encompass the whole of the story's universe, will be enough to carry the reader all the way through to the end. It takes, for lack of a better word, chutzpah.
The crazier the conceit, the more chutzpah you need to see it through. Sometimes the only way to do this is to play it all with a straight face, as Christopher Priest did in Inverted World (which, in a good example of "popular" and "literary" fiction shaking hands, has since been reissued by none other than the New York Review of Books). By doing that, he took what might have been an unsupportable concept and not only made it supportable but turned it into a larger allegory for human progress as a one-way street.
Sometimes you just need to be as crazy as the material demands, as Brian Aldiss did with Eighty-Minute Hour. There, he made no excuses for what was happening, but sustained it all by being massively entertaining. The consistency was in the tone and the approach, which was uniformly comic and tongue-in-cheek.
Angela Carter's Nights at the Circus — a splendid book which doesn't get nearly enough press, by the way — walks somewhere between the two. There is one major element which requires suspension of disbelief, and then a whole slew of other ones — but by the end of the story, the whole thing has become a treatise on the nature of suspending one's disbelief and building trust with an audience, both the ones in the book and the ones outside it flipping the pages. I have my own problems with the story being a little too overstuffed and headless for its own good, but they stand entirely apart from how entertaining and imaginative it is — and how the whole thing also works as a sly jab at the way men project their fantasies about women onto the real things, often to the detriment of both.
Consistency, then, isn't just a matter of laying down conceptual rules. It's also a matter of tone and approach, and those are things which readers respond at least as strongly to, if not more so, than the what-if that fuels the story.