Look at the Mycin dialogues again. Are Mycin's answers to how and why questions adequate? All it does is to feed back the rules that it's currently investigating (for why questions) or has investigated (for how questions). As you should see from the backward chaining examples, this is easy; but it is not good enough.
Until about 1983, it was generally believed that this kind of explanation was perfectly adequate. A great advantage of expert systems was said to be their ability to automatically generate explanations suitable for novices, without any extra work by the knowledge engineer. But these little nuggets of detail are useless to someone who doesn't know the underlying structure of a subject. What's needed for novices is (?) an overview, concentrating on general principles. A system like Mycin can't provide this because the only knowledge it has are its rules; there's no information on why the rules are what they are, or when they are appropriate. For example, in many expert systems, the order in which rules are used affects the conclusion. But there's no explicit knowledge saying why the rules are in that order. Maybe the author ordered them in order of specificity, or execution time. In the terms of Goodall pp 100-101, the conflict-resolution strategy is not explicit.
There has been research on making strategies explicit and incorporating them in explanations. For example, Neomycin. See Strategic explanations for a diagnostic consultation system by Hasling, Clancey and Rennels. In Developments in expert systems edited by Coombs (PSY KH:C 078).