Look, let's face it: Arbor Software commissioned Codd's "white paper" defining OLAP -- does Codd otherwise have research interests in this area? (serious question) -- and other vendors jumped on the bandwagon because the market picked up on this new grabbag software class. Codd's definitions muddle logical and physical models, and the skewed interpretation of his rules (e.g., that "unlimited dimensions" means that unlimited dimensions can be defined and not that an object can have arbitrarily-many dimensions) and the numerous additional rules proposed by vendors and analysts support how inadequate his rules are. Plus the "analytic" part of most OLAP products is a relatively small slice of what analysis really means.
I wish the term "OLAP" would die. Other, existing terms like "data warehousing" and "decision support" are more precise, and most of what is now called "OLAP" could be more accurately labeled "dimensional analysis." If you don't like the "dimensional" because it doesn't provide adequate contrast (and name similarity) for this class of products to OLTP, try Philippe Kahn's OLCP -- online complex processing -- which doesn't falsely imply that true analysis is going on.
P.S. Kahn coined the term OLCP to describe Interbase, Borland's server RDBMS, which manages concurrent use with versioning instead of locking and which was one of the first commercial RDBMSes to offer advanced features like user-defined functions that are of primary interest outside transaction-oriented applications.