How long does it take to develop one hour of training?

Examines the results of a US survey on the time taken to create an hour of live e-learning.

People often ask how long it takes to develop one hour of self-study e-learning. The answers vary wildly, from under 50 hours to more than 300, depending on the amount of research that is needed, the complexity of the interactions, the richness of the media, the capabilities of the authoring tool, and the experience of the designer. These figures nearly always surprise people, because they wouldn’t normally spend anywhere near this time developing for the classroom. However, because they have to stand alone, self-study materials are notoriously hard to develop and they can only therefore make economic sense when there’s a reasonably large audience of users. The estimates are also open to question on the basis that self-pacing is, by definition, variable – what’s one hour for one learner, is 20 minutes for a second, and 2 hours for a third.
However, with live online learning, the concept of ‘one hour of e-learning’ really does make sense. An hour is an hour is an hour. That’s why I was interested to read the analysis by Karl Kapp (see Time to Develop One Hour of Training):
“In 2003, the low estimate for developing one hour of instructor-led, web-based training delivery (using software such as Centra, Adobe Connect, or WebEx) was 30 hours and the high estimate was 80 hours. In 2009, the low estimate is 49 and the high estimate 89. Both higher. Is it taking us longer to develop e-learning than it did six years ago?”
These figures are low compared with self-paced e-learning but higher than I would have expected. I can’t quite see why it takes 1-2 working weeks to assemble a really good hour of training. Am I missing something here? What’s your experience?