Your coherence and trustability, on the eyes of others, can rely, and will tend to rely, in what you publish (especially in intellectual affairs and environments). In a world of works obtained by collaboration with others, co-authorship is authorship. Still, in the long term, you can build yourself a clear trustability anyway. Everyone knows that when working in collaboration, perfect agreement is scarce.
It is unethical (that depends of your ethical criteria!) to publish unfounded research, for responsability and compromise with society. However, depending on the type of paper, some sections can be a matter of interpretations and weighting of data presented. So I do not find disagreement unethical, but erratic behavior rather than straight-forwardness and honesty are probably improper in intellectual environments working in publishable and potentially influential material. Talking about ethics: everyone has free will, so you may not be able to prevent publication (if that was necessary), but looking for discussion and agreement will probably lead to better results.
1. Expose your opinions clearly and request an answer (some people omit their opinions; requesting explicit answers is increasingly necessary).
2. Be proactive: suggest solutions that would solve the problems you expose.
3. Debate to find agreement if there is no agreement after the first step.
4. If no previous steps worked sufficiently for your personal criteria, evaluate whether you can afford not publishing it or you actually win by not publishing that.
5. If found convenient, politely indicate that you would prefer to withdraw from the authorship of the paper if certain considerations are not stated differently.
I would never delay the straight-forward approach: waiting until review will make the potential future corrections and agreement impossible.