What is it that makes delivering a good, clear, informative lecture so challenging?
It is challenging because learning is challenging. The things that one could easily convey to a very general audience simply by being clear and informative are by and large not the things (or certainly, not all the things) that one learns in university-level courses.
What makes a good lecturer is a great question, but a bit broad for a site like this. If you have a group of students / faculty colleagues, I highly recommend throwing this out to the group: break the question up into some directed subquestions, give everyone the questions, and then after a week or so meet to discuss their answers. I actually did something like this recently (the topic was "successful math talks"), and what ensued was entertaining and enlightening.
Here let me (still very superficially) try to address your observation that even Nobel Laureates need not be good lecturers. Again I will break this up into a few subquestions.
1) Do you expect a Nobel Laureate to be a better lecturer than a faculty member with a less exceptional research profile? Why or why not?
2) Do you think that Nobel Laureates would be especially good or bad at giving certain kinds of lectures, or lectures to certain kinds of audiences? Why?
I think everyone can get a turn at answering these questions. (Probably not here: the site is not designed for that.) Let me take a crack at the first one:
One can argue that Nobelists ought to be on average very good lecturers. First of all (contrary to what some people would like to think), intelligence and acumen in one domain is positively correlated with intelligence and acumen in another domain, and the correlation increases to perfect as the two domains converge. If you are a professor of X, then researching X and teaching X are your two main duties. They are different, but I can isolate a common variable. The best way to be a bad teacher is to have a poor understanding of your subject, and Nobelists must be the least at risk for that. The people who have not thoroughly mastered their field, especially at the level of coursework, are in my experience essentially never the people who are making the cutting edge breakthroughs.
Is it possible that when a true luminary gives a lecture, we evaluate them with that high standard in mind? I often tell the story of a colloquium I saw given by a Fields Medalist ("the mathematical equivalent of the Nobel Prize"). By chance I had read about his work, relatively casually, about a month before his lecture. I was profoundly disappointed by his lecture because it was merely clear and informative. The information he conveyed was almost exactly the information that I had read before. However, when I read about I found it incredibly exciting and impressive. In person he did not convey any of this: it was just a recitation of "In 1982, I proved this theorem; two years later I proved the following improvement". I felt afterwards that if for some (totally counterfactual) reason I had given the talk instead, I would have done a better job, because I would have been so enthusiastic. Mine is a sincere reaction, but you see that I was coming in with very high standards, and the idea that I could have done better ought to be construed as a description of the psychology of my reaction: taken literally, it seems rather unlikely.
One can also argue that it is not so surprising if Nobelists give lectures that are merely okay or actually not as good as what other faculty are doing. (Again it depends a lot on what kind of lecture we're talking about, which I am omitting for now.) Let's go back to what I said before: being a professor of X involves researching X and teaching X. Yes, these skills are positively correlated. But they also compete for our time. If you are a subject area expert with some teaching experience, then you can probably deliver a decent lecture on X with a moderate amount of preparation. But maybe moderate is more preparation than you want or feel that you can spare. If you don't prepare at all, then no matter how brilliant you are you are probably going to give a lecture which the audience will regard as being rough -- maybe too rough. If you want to give a better lecture then you probably have to prepare more. Now there are a lot of people in university environments who are spending much more time and thought preparing their lectures. It is not a zero-sum phenomenon because no one tells you exactly how many hours you should spend total in any given week. There are a lot of leading researchers who are clearly putting substantial time into their lectures. However, Nobelists are the extreme case: you are selecting people for which the community as a whole feels that their research is much more valuable than their teaching.
Taking these two things together: I would expect a Nobelist to give lectures that are excellent in some ways and below average in others. By the way, what makes a good lecture is a many-dimensional space, and many is the time that a friend and I have walked out of the same talk and discovered that one of us loved it and the other hated it. But if you're judging a talk by the types of things that go into undergraduate student evaluations: I think you're mixing together a lot of highs and lows and thus one should expect the results to be pretty much all over the place.