I think Shelah's book is very good, but it is very difficult to read, in part intrinsically because of its subject matter, and in part because it was written by Shelah. Recently, Richard Elwes reminded us of the review by Lascar (I do not agree with portions of the review, but it gives you a good idea of what you are up against):
Unfortunately, the book is very difficult to read. This is undoubtedly due in part to the difficulty of the topic itself, but it also has to do with the way the book has been written. The reader should be warned about the numerous misprints and inaccuracies he will have to detect and the time he will need to work out the numerous "easy'' or "left to the reader'' proofs. He will have to understand that statements made under impossibly complicated hypotheses for the sake of generality, and often in an axiomatic setting, may conceal, as a special case, a simple and important fact. Also, he should not expect a rigorous structure: notions are usually introduced when needed, and it will be up to him to guess how important they may be in the rest of the book.
All of these features do not make for easy reading, and we shall certainly not advise anyone (except perhaps the author himself) to "devote himself to reading and solving the exercises till he knows the book by heart'' (cf. Introduction). We do not recommend it as a textbook, and it will be difficult to use it as a reference book, considering how hard it is to find any particular result. But we do think that nobody involved in research in model theory can avoid studying it. We are unable to recommend any particular angle of attack. We did find the opening remarks in each chapter rather helpful (with the exception of Chapter III).
For modern work in the subject, you probably want to read eventually about Geometric Stability Theory, as in Pillay's book. An introduction to stability that would better prepare you for this would probably be best. Pillay's introductory book on Stability Theory is certainly an option. Other suggestions are Marker's book, Poizat's and Hodges's. The truth is, recently there have been quite a few nice texts published, so it should not be difficult to find a good text with which to start.
(But I would still suggest to have Shelah's book as a reference and to consult it as you learn more of the theory.)
On a vein different from the algebraic geometric approach, you may be interested in how set theory interacts with the subject. Shelah's book has plenty of this while constructing non-isomorphic models, but there is significantly more in this direction, in the context of abstract model theory and abstract elementary classes. I would recommend Grossberg's text as an introduction.
