That's a problem I've been having with integration between mountains. It's a major reason I haven't been posting much on this lately(a little bit of my trial and ERROR process goes a long way...). I like the valleys in your modified version, but I think the mountains suffer. Although, those are terrific as hills or lower mountains.

Curves is good for small adjustments, in my opinion, but really significant curves force too much regularity on the landscape. My opinion.

I've been adding fractal noise with Clouds and Difference Clouds to try to disguise some of the problems, but, as I think Redrobes has noticed, this produces a lot of sinkholes, which just ruins the drainage. This is important if your trying to simulate decent erosion in less than a lifetime. And, at best, you still end up with something that looks vaguely like glacial u-shaped valleys.

One experiment I'd like to try would be to take the heightfields resulting from both of our methods, scale them in Levels and add them together. That could either(hopefully) integrate the best aspects of both methods or (unfortunately) amplify the worst aspects of each.

I'll try that when I get a chance. Tell you how it works out...