At National Lab, where bean counting was becoming more and more important, the impact factor trumped all. All raises and promotions had to be justified to non-scientists and/or former researchers in a wide array of fields. To make things easy, they would rank papers by impact factor, regardless of field. This is hugely distorting, since a great specialty journal in physics can have a smaller impact factor than a decent specialty journal in biomedical research (just due to the number of researchers in the sub-field). To get ahead, many people would find the journal with the highest impact factor that would publish the research.
My preferred style is to find a journal where my target audience is likely to read it. So if I am doing physics-type measurements on a protein, I would prefer to put it in a physics-type journal (where others might pick up on it and find it useful) rather than a protein-type journal where most readers could care less about this physics-type measurement, even if the impact factor is lower. Even though my bean-counting impact might be lower, I am counting on the actual impact being higher. What is the point of a high-impact publication if no one reads it? That said, when DrugMonkeyasked people about their citation practices, many commenters said they prefer to cite papers from higher impact journals, so clearly impact factor must be balanced with target audience.
This type of selection is much easier when I am intimately familiar with the sub-field. In that case, I know well which journals I read and cite, and can select accordingly. In this particular case, I don't have any knowledgeable colleagues or collaborators I can talk to, but that is generally my first stop in deciding where. How do you all pick where to submit your work (especially work in a new area for you)?