The deep coastal canyons look a lot like fjords (glacial canyons carved below sea level), which is why I suggested the coast of Norway as an example. Being off by a factor of two using something like Wilbur is not too bad. Making it controllable is the hardest part of something like this task and I'm not sure what you were aiming for when you started. A good self-challenge (even if it's only a private one) is to generate a rough map with the general features that a customer is likely to want (coastline shaped like so, high mountains here, low mountains here, plains here, rivers along there, and so on). Make a high-level map using that rough map and then a couple of zoomed-in areas from that first generated map. If you can preserve the required features and maintain plausibility, then you'll have something. Having said that, you're doing fairly well with generating terrain so far.

An important aspect of cartography is managing expectations. If you had started out with a scale bar on the image to indicate that the map is intended to be 2750km across, then you'd have set expectations and offered hints as to the scale of individual features. Wilbur is particularly hard to do plausible things with pixels much larger than a hundred meters or so because of how the algorithms work. Fortunately, another important aspect of cartography is generalization: only show what's important to the map. Pushing unimportant details to the background while keeping the important things uncluttered is really hard, no matter what tools you're working with.

I can't recommend enough the idea of looking at the works of others. http://shadedrelief.com/ is an excellent starting point for getting some how-to ideas. Decide if you want to go with straight hypsometric shading or if you want to look at other features as well such as moisture. I've said this before (and in no way am I trying to make light of your efforts): generating terrain is the easy part of making a map. Making the customer happy is the hard part.