The problem with using it to impart a method of thinking in my mind is that the way of thinking (and the patterns themselves) really have to be earned. I think it's possible for a book to do a fair job at imparting that wisdom, but GOF is setup terribly for it. It's sort of like picking up a recipe book full of classical french dishes without learning about french cooking first. It's too easy to misunderstand the specific values and compromises of specific 'ingredients' and undo the benefit the pattern could provide altogether.
I've wondered before if GOF could be rethunk with Refactoring in mind. Starting with code smells is exactly right. The book should focus more on identifying problems with an approach, because that's the hard part. Solutions are much easier.
That's why I like PoEAA - Fowler is better writer than the GOF so his books do a better job of bringing the ideas across.
new $className();I really really hate singletons, visitors, factories, and all the lot, but I'm glad I can understand the intent of these things.
I think being disillusioned with this book is a good stepping stone for devs, it felt like one to me.
As far as coding skill, I agree about Refactoring. Design Patterns is just a grab-bag of tools that are useful in some situations and harmful if misused or abused. Refactoring is a set of skills that are globally useful and are married to equally important models, terminology, and advice. Studying refactoring gives devs a much stronger ability to talk about, study, understand, reason about, and modify code. I'd much rather work with someone who was an expert in refactoring than someone who was an expert in design patterns.
Edit: also, Refactoring concentrates on changing existing code vs. greenfield development, which tends to be far more useful in practice.