Mental illnesses are generally scientifically determined by and pathologically defined by the functioning level of the case in question, being the functioning that is defined by ALL parts of a normal life, i.e., social, work, personal, mental and physical.
Thus, any absurd notion that mental illnesses are in a haze is firmly removed.
The next problem is really the diagnostication. Not only are the the categories unclear/definitely undecided by consensus, but the approach to the patients is not entirely set by procedure either. I have this fourfold suggestion:
1. Cognition of patient's language. You can make good manuals for clues to look for in the patients.
2. Patient's behaviour. This is really the brain-child of B. F. Skinner and is still in good use, although a little more intelligently, like mimicry, possibly by computer pattern recognition.
3. The classic questionnaires incl. (the rather unserious) Rorschach test.
4. Patient's self-reports and general reports about the patient by close friends, family, etc.
Not only are these good, but you can still add the metabolism test from blood sample and new approaches by (f)MRI.
By this, diagnostication should be seen as 100% and there should be little margin of error unless the mental illness is in its very early stages!
Cheers!
(more later...)