One common misperception, I think, is that the traditional, content based approach works. It kind of does, but I think there are a couple of factors at work:

• First, a great deal of traditional training is at the lower end of the explicit/tacit scale. Straightforward procedure dominates: this is how to do X, and X isn’t something like “analyze data-integration needs for a Fortune 100 company.”
• Although planned performance support is pretty thin on the ground, there’s ad-hoc support. First, a lot of systems won’t let you do what you shouldn’t do, so you “learn” not to try that.
• Heyjoe support abounds.
  

What this means is that there’s a lot of after-the-fact, improvisational, figure-it-out stuff going on that makes traditional training look better than it is. Abstraction of results continues as you go up the organizational hierarchy, so the top brass doesn’t often ask about the gaps between training activity and on-the-job results.

Another good point – there is simply too much information to be assimilated. I believe maximum workplace performance occurs when you understand who you are as an individual (strengths, personality traits, values, etc), you are in the right job (good fit), and then you are plugged into the people and systems around you. That’s when learning happens without realizing it.