Say you are a condensed matter physicist, and you have submitted an article for publication to the prestigious (well, kind of) Physical Review Letters (PRL). You did so because you are objectively convinced of the novelty, importance and broad interest of the results that you and your collaborators have obtained. Suppose that the reviewing process is lengthy, difficult, time-consuming and aggravating, and ends with the rejection of your manuscript. It was a close call — several rounds of reviews were needed, and the manuscript had the support of one or more referees, but in the end the Editor decided not to accept it. This is actually a fairly common scenario.
Often times, the Editor of Physical Review B (PRB, the condensed matter section of Physical Review) will offer you to publish your article therein, typically without the need for further reviews.
I have the feeling that there are many articles published in the specialized sections of Physical Review which had initially been submitted to PRL. Some of them garner a substantial number of citations, possibly contributing significantly the the relatively small difference in Impact Factor (IF) between the specialized sections of Physical Review and PRL .
Authors accept the offer of the Editor of PRB (in the above example), even though PRB is not as prestigious as PRL, primarily because they are sick and tired of dealing with reviewers and are eager to see their work published, in one form or another. The present lack of other physics journals that can compete with PRL in terms of IF makes it scarcely compelling for authors to submit elsewhere articles rejected by PRL.
However, I am becoming convinced that we should all make an effort not to accept the “wooden spoon”, the consolation prize described above, namely the quick and relatively effortless publication in a specialized section of Physical Review. Instead, the authors should bite the bullet and submit their articles to a different journal, even at the cost of a prolonged review  for the following reasons:
1) The fact that articles that are “borderline” for acceptance in PRL, for which the decisions are difficult, end up in all likelihood in some specialized section of Physical Review, constitutes for the Editor of PRL a powerful incentive to reject, because the article will be published in Physical Review anyway. Things might be different if the Editor of PRL worked under the assumption that the article may well be lost to a competing journal.
2) By submitting to a competing journal the same article, an article which in our expectations may receive a respectable number of citations, we contribute to enhancing the IF of that journal, in the long run providing all of us with an alternative to PRL for submission of our most significant pieces of work. I do believe that our community stands to benefit from having alternatives to PRL as upscale publication venue.
 Yes, I am suggesting that wrong editorial decisions may be an important, often neglected cause of PRL’s declining IF.