Obviously there are rather a lot of people who have used MetaEdit+: from projects that have been going for 15 years with hundreds of modelers, to the rapidly growing mass of intro licence users.
Maybe the best place to look for in-depth experience reports are the industrial cases that have been published in various journals, books and conferences, many with empirical evidence backing up the 5-10x productivity gains - something I've not seen for any other tool. I've just published a list of some of those on my blog, titled "Domain-Specific Modeling: MDD that works".
There is an interesting small contradiction (or inaccuracy) in MetaCASE's claims, Steve.
In your blog post, the second graph shows a comparison of the cost (I assume time) of different approaches. You claim that modeling the first product with a DSM tool takes the same time as doing it by manual coding, which I am sure is not correct, by my own experience. This is also backed up by Laurent Safa's paper at DSM'07, one of your DSM reference cases (http://www.dsmforum.org/events/DSM07/papers/safa.pdf):
"We found that each page standard took us around 3 man-days to implement the metamodel, the visual notation and the code generation. When compared to an estimated development cost of 0.5 man-day per page, we assumed that any page that occurred 6 times or more was worth metamodeling."
Don't take me wrong. I truly believe in DSM and I think it is the way to go. However I think that the cost of modeling the first product is higher than the traditional manual coding approach.
The second graph doesn't show the cost or time of different approaches, it shows for a constant time how many products you can make with those approaches. Or as the blog text puts it: "Imagine that you can build one product in a certain time by coding. How much more can you build if you apply various kinds of modeling?"
So if in you can build one product with manual coding in a given amount of time, in that same amount of time you can build about 0.85 products with UML, or 1.2-1.3 with MDA, or 5-10 with DSM.
For the second graph we're looking at the "steady state" once you're up to speed: you've learned Java, bought the MDA tool, built your DSM tool etc. As the blog goes on to show in the third graph, there's a start-up cost associated with setting things up for DSM in your domain. I don't try to guess the cost of learning Java, UML or MDA, or buying those tools - I don't know the figures, and for at least Java and UML I guess it's already happened for most developers. Buying a DSM tool is comparable to buying an MDA tool, so with DSM the extra cost is creating the language and generators. That took 3-16 person days in these cases, so isn't huge.
In some cases DSM pays back in the first product (e.g. Polar); in others it takes 2-3 products. That's really more dependent on what constitutes a "product" in each domain: the key thing is DSM paid back within the first four days of modeling in all these cases!