Sorry to have missed most of this conversation, given its a huge area of interest for me. Ironically, was overseas chairing a symposium peripherally related to this exact topic. Random series of thoughts:
1) Operational definitions of self-medication are important as the term has broadened substantially and is no longer just one "theory". The original notion was that individuals would develop dependence on substances as a method of "self-medicating" other psychiatric disorders. PSYDR is right that evidence for this is VERY mixed and far from conclusive (arguably leaning in the other direction). In part, I think its because the notion is somewhat misguided in the first place and based on our somewhat archaic diagnostic system that we know is wrong but haven't yet come up with anything better. Healthy populations could self-medicate normal fluctuations in mood. Individuals with disorders could self-medicate "normal" aspects of their disorder. Disentangling normalizing/enhancing effects of drugs of abuse from withdrawal relief is unbelievably challenging.
2) Self-medicating some symptoms may or may not change the diagnosis. Again - archaic, clunky diagnostic system.
3) Clinically, I'd say the bulk of evidence leans towards "treat what is there - who cares what is 'primary'" anyways in the overwhelming majority of cases. You rarely see iatrogenic effects and usually see benefits from treatments addressing multiple topics. Heck, despite profound resistance to smoking cessation in addiction settings for years ("focus on recovery and worry about smoking later"), we're seeing growing evidence that it actually improves abstinence from drugs/alcohol. Studies are typically non-experimental though, so jury is ultimately still out on that one. I'm not sure much is gained clinically from thinking about self-medication, though I think it helps frame our understanding of the problems and is beneficial for understanding the reinforcing effects of drugs and potential future research.
4) Keep in mind the things people may be self-medicating can be known consequences, but not necessarily DSM criteria for the disorders in question or even easily measured (e.g. sensory gating in schizophrenia).
5) Also keep in mind that any "self-medicating" effects of drugs of abuse may be indirect, which means there are potential moderators, confounds, etc. to account for. For instance, both alcohol (Curtin et al., 2001) and nicotine (Kassel & Shiffman, 1997) may NOT directly impact mood - but do so indirectly via impact on cognitive control. This is going to be tricky to measure and means any self-medication success would likely be context-dependent. Either way - drugs of abuse are pharmacologically dirty, we can't expect effects to be clean. Any efforts to identify a single transmitter system are likely to fail.
6) Nicotine is particularly tricky, since its reinforcing effects are wacky. Its extremely addictive in the form of cigarettes in the general population of smokers. Its a terribly weak primary reinforcer and getting rats to self-administer it is a tremendous PITA compared to something like cocaine. Its action may be as a secondary reinforcer (see work by Donny, Caggiula, etc.) but we're struggling to measure that well in humans. The 2000 other additives in cigarettes certainly complicate things. Some evidence menthol and other factors may interact with nicotine, but aren't typically examined in the basic pharmacology studies. Nonetheless, environmental/social factors may play a stronger role than the drug itself when it comes to smoking.
7)
IMO, Read's work uses an impressive methodology in a field that typically reports correlation or comorbidity. But she indicates alcohol use is a trait vulnerability.
With all due respect (and assuming we're talking about the same Read), I know her and I'm extraordinarily doubtful she would describe alcohol dependence (let alone alcohol use) as strictly a trait vulnerability, though some individual studies of hers certainly provide support for trait vulnerabilities contributing to substance use. She'd likely argue its an interaction of traits, experiences/environment and a multitude of other factors. She's not one for oversimplifications and in this field - picking "one" theory to hang your hat on is almost always going to be an oversimplification.