I'll restrict my attention to analysis and algebra. (I've never met a topology book I liked much.)
I think analysis is inherently grungier than algebra, and I've never found a flawless book. Either we get a clean, abstract approach, where the author doesn't provide much motivation and the exposition is somewhat detached from anything classical/concrete (Rudin), or we get a grungy, let's-get-our-hands-dirty treatment where the proofs are trick-free but also tedious and you start wishing for a bit of that abstraction to clean up the arguments (say, Spivak, but at least his exposition is very insightful). Sometimes, we get abstraction without the beauty (Folland) or grunge without much insight (most analysis books).
A short list of analysis books that I think are well written, insightful, and strike a decent balance between the abstract and the concrete:
* Stein and Shakarchi's four volumes
* Bruckner/Thomson, both the undergrad and grad books
* Carothers
* Bartle, Elements of Real Analysis and his little Lebesgue book
* Berberian, Fundamentals of Real Analysis
For algebra, it's certainly possible to write unclearly or unpleasantly (Lang, Hungerford, anything with more commutative diagrams than text), but the subject seems inherently cleaner and less in need of being tied to anything "concrete". Or maybe that's just a matter of taste on my part. I don't particularly care if there is any non-mathematical application whatsoever of group theory, I love it for its own sake.
The three best-written books I've encountered on algebraic subjects:
* Rotman, Advanced Modern Algebra
* Roman [not Rotman], Advanced Linear Algebra
* Isaacs, Finite Group Theory
