This scientific idealization also disregards the deeply political and decidedly non-scientific nature of many aspects of accounting and stands at odds with the emphasis on experience, judgment, and artistry used to justify CPAs' role in financial regulation, as formalized in the 1930s in the wake of another financial crisis. Building on accounting's pragmatist roots and contemporary legal literature on accounting, the article concludes that substantive reforms should better incorporate elements of judgment and subjectivity by requiring increased disclosure and transparency, but not prescribing specific accounting treatments.
You can read the whole paper in order to get the idea. It seems to echo Amar Bhide's A Call for Judgment.
I doubt that one would want to imply that judgment will succeed where models or rules fail. Rather, the important point is that precision is elusive. Instead of a number, it may be better to report a range. In fact, even a range can be misleading, because there may be scenarios (or interpretations) under which results will fall outside the range. So there is something to be said for reporting alternative scenarios or interpretations.
For tax accounting, I would think that we need strict rules. It won't do to say that I owe something in the "range" of $X + or - $m or to say that there is an alternative interpretation under which I owe $Y.
On the issue of whether accounting flaws cause or are merely exposed by a financial crisis (my question originally), Green writes in an email,
abuses of whatever accounting method is currently in fashion are cyclical, and a period of abuse triggers a reaction back to the other -- from historical cost to fair value or (currently) vice versa. But the rightness or wrongness of accounting per se is probably not cyclical, except insofar as auditors possibly let up their guard in the trough between crises.