Hi EricVT,
I've heard a similar comment at a panel discussion a few years ago - I highly suspect from the same person.
Now that I have some experience reviewing papers I don't doubt that this is the case across most science fields... at least for journals that you don't have to pay to publish in.
For one, the acceptance rate for the major journals in my (our) field is only around 50-60%. That means that only half of what gets submitted is *eventually* published. I suspect that the percentage of papers that only need a single round of review with minor corrections are much lower than that - probably in the 5-10 % ballpark.
On top of that the peer-review process can be highly subjective and if you consider the small pool of very good initial submission papers,l I suspect that in many cases where the reviewer can't identify any direct problems with the science he or she is left in a rather awkward position. What if there is something wrong and you've missed it? Will anyone believe that you actually read the work if you don't have anything to say about it? Is it even possible to write something that is so air tight you can't find any problems with it?
As a result the reviewers, I believe, come back with more subjective comments for improvement to the work - things such as increasing or decreasing the length of certain sections, or rewording, or requests that certain figures get replotted, or that certain tanget subjects be discussed or commented on, etc.
But this could be field-specific too. I would guess that in some of the large scale sub-atomic groups there is so much internal drafting and re-drafting that by the time a paper is submitted to a journal it's been critically assessed so much that it's nearly impossible to find something objective wrong with it.