There is an endless amount of Lisps because the core to get one working is one page of code. So many people make one as a pet project. Aside from the small unused versions, the rest of the dialects have a lot of differences between them, some of which would be difficult to implement in an existing implementation. Common Lisp is a fairly complex specification to achieve, so few even approach it, and of those that do most aren't fully compliant.
Getting a Lisp to run efficiently and effectively is more difficult than just writing an interpreter. I consider work on this problem to still be incomplete. For example, it should be possible to implement generic functions even more efficiently in Common Lisp implementations than is currently the case, in a way that preserves the ability to redefine the functions (and the classes they work with) dynamically. Robert Strandh has been working on general mechanisms supporting this and other improvements.
