It comes across listening to fellow grad students/post-docs at conferences, at graduate school, SnobU, etc. that different labs take very different approaches to getting work published. Some labs think in terms of the minimal publishable unit (MPU) and set up their experiments accordingly when starting a project. This can lead to rapid publications though at the expense of well supported conclusions but if you can spin it well you can get away with it. The main competitor of my current PI is like that. Very well written work comes out of his/her lab with an exciting spin that the data doesn't quite support. The other extreme is doing experiments and letting the data take you where it leads. This is the ideal of the scientific method that is taught from primary school onwards. You have a hypothesis, you design experiments to test it and based on your results you refine/alter your hypothesis and design more experiments. Once you have enough support you write up and publish. The conclusions are usually very well supported. The danger is getting scooped by the MPUers in the meantime on certain parts of your project, lessening the impact of your work. In addition you may have to go back and redo certain things to generate figures/tables that would fit the journal you are going to submit to though not adding any new results, just showing the ones you already have in a different way.
Most labs are not either but on spectrum between the two extremes. My lab is such a lab. We ask questions do initial experiments, generate results, ask more questions, etc. but at each point we sit down and talk with our PI about what will be publishable, thinking about where we would publish our work. As we generate more and more data we sit down and critically think about our projects, trying to think what a reviewer might say. Not just thinking of the experiments to be done but also how to present them and what journal we should submit to. Usually this speeds up our ability to publish while still having well supported conclusions. When the PI was having health problems & before I joined, the lab did publish a few MPU type papers. The conclusions of one were completely wrong. To my PI's credit, he/she went through the task of showing where we went wrong and doing the necessary work to reach a well supported conclusion.
Approaching writing a manuscript is also crucial. I know many labs try to shoot for the highest impact journal they can. The mindset being "it doesn't hurt to try". Actually, it does! Writing papers takes time. A Science paper is written and presented differently than a Biochemistry paper for example (not to mention the type of data tends to be different). If the paper is rejected, it means you have to rewrite the paper for a new journal. This resets the clock on the review process, further delaying your work being published. The rewriting also tends to take away time from designing/doing new experiments and/or having a life outside of lab. Not to mention, rejections can be demoralizing. On top of that, there is a good chance the reviewers will be the same. A couple rounds of submissions and these reviewers can tire of work, making it even harder to publish in the lower impact journals.
If you are in a competitive field, shooting too high can cost you. In the time that you are trying out journals, your competitors could publish similar work, scooping you while also having the time to start on the next round of experiments, setting you further back.
It is worth it therefore to sit down and really think about your work before writing. What are your main points you are making? What type of data do you have to support those points? What sort of impact does it make? What general scientific interest would it generate? Solicit the opinion of others. The work you are doing should be exciting and wonderful to you. That passion helps to motivate you to put the effort that is necessary but it does tend to make us less than objective about our own work. Listen to those you respect so you have a proper perspective. This is also very useful for those that tend to beat themselves up; don't undersell/shortchange your results and ideas either. Once you have a good idea what journal would accept your work, write a your manuscript accordingly. Have it read and edited by multiple people, multiple times. Make sure they are more critical than a reviewer (in a constructive way). Typically this approach leads to papers that are accepted with minimal revisions, saving you time to do more experiments and to have a life. The most successful PIs I have seen are the ones who learn this lesson early in their careers saving them time, money and maybe a little of their sanity.
Dr. Mom discusses impact scores further and where to submit your work in her wonderful Writing Your First Paper series.