As a preface, although your concerns about the validity of the field of chiropractic are entirely warranted, it appears that Dr. Stuart M. McGill is indeed qualified. He was a former professor at the University of Waterloo, holding Ph.D., Masters, and Bachelors degrees in Kinesiology (Biomechanics), Kinanthropology (Biomechanics), and Physical and Health Education, respectively. So whilst there appears to be no other literature to corroborate his findings, without performing the bio-mechanical calculations ourselves, it is reasonable for us to accept his claims.
It seems fairly clear from McGill's tone that he does not support the prescription of sit-ups: “...performing sit-ups using bent knees or straight legs is probably not as important as the issue of whether to prescribe sit-ups at all!” However, he states only that “Compressive loads in excess of 3000 N [~300 kg] certainly raises questions of safety for some patients.” The same thing can be said about all exercise. Thus, the question really becomes one of degree. We might reasonably ask, prescribe to whom? And for what reason? McGill's subsequent analysis suggests strongly that he is evaluating sit-ups as an exercise for the general (unconditioned) population and as an abdominal exercise. Furthermore, he is clearly making assumptions about the way in which they are to be performed. And from his perspective, I believe, he puts forward a very good case against sit-ups.
However, that does not imply that sit-ups are inherently dangerous or harmful—only that they are potentially dangerous for a certain subsection of the population, and that there are objectively better choices of exercise for abdominal development. So in order to assess the former question, we should again examine the literature.
Virtually all of the existing data surrounding the link between sit-ups and injury come from the military. One oft-cited study, for example, revealed that 56% of the injuries recorded from a sample of 1,532 soldiers performing the biannual Army Physical Fitness Test (APFT) were from sit-ups. This evidence certainly appears rather damning, especially since running ‘only’ accounted for 32%. However, when we consider those data in their context, the danger of sit-ups becomes rather less clear. That study revealed an injury rate of 7.6%, or one in thirteen—a staggering rate of injury for a test consisting of just three disciplines (push-up, sit-up, and a 2-mile run)! But given the demands of the military, and the seasonal variation in training and conditioning that has been observed, we can safely assume that this observation represents an exception. Indeed, it is notable that the authors of the study concluded that “The push-up, sit-up, and run events of the APFT do not pose a considerable acute injury risk to active duty soldiers.” Clearly the military is willing to accept greater risks than most of us might.
Given their ubiquity in physical training, if sit-ups were inherently dangerous, we would expect a plethora of hard data demonstrating that fact in the literature. But no such data seem to exist. We can reasonably conclude, therefore, that they are not inherently harmful or dangerous.
It is certainly notable that compressive loads may exceed recommendations for the normal population. We should therefore objectively class sit-ups as a moderately difficult exercise, and one appropriate only to healthy-weight athletes with some level of conditioning. And we should further take measures to prescribe or perform technique that would minimise spinal compressive loads and consequent injury—in particular maintaining ‘neutral’ curvature of the vertebral column throughout the movement, both at the lumbar and cervical spine (that is, without ‘crunching’ or pulling the head forward).
On a final note, the sit-up should be understood to be an exercise for the hip flexors, with the abdominal muscles functioning only to support the spine isometrically. This is clear from McGill's article, which summarised the relative contribution of the different contributing muscle groups.
I hope that is helpful.
