Eileen,
Trying to get any decent reporting system out of 400 tables isn’t fun. I bet it is also highly normalized (Chasm traps, Join Problems, Contexts errors). That type of Transactional system is real good for getting data into it (Nice fast Insert statements), but is horrible at getting data out of it, e.g. Reporting and/or building BO Universes for reporting.
The transactional system that I’m working with was/is about the same size and they had the same problem. My clients solution was to hire the firm that I work for and we built a classic Kimball designed Star Schema (Check out “The Data Warehouse Toolkit” by Ralph Kimball). So now we have 6 Fact tables broke out by business needs, holding about 60 plus measures. We have two levels of granularity, one at a summary level and one at a detail level. E.G. the summary tables have about 2.5 million records and the details have 10~200 million records. The Fact tables are then surrounded by a core group of 8 dimensions - Time - Product - Promotion - Status - Customer - Sales Org. - etc…
From the BO Designer perspective, I take the 9 Core Dimensions and create a CORE.UNV. This allows me to keep all the core dimensions in one universe. None of the dimensions are joined. Any filters and Alaising is also applied here. The main benefit is that I when I make one change to an object in the CORE.UNV it is automatically cascaded to all the derived universes.
I then group the Fact tables by Business Need and link them to the Core Dimensions. Here is the Example:
Core Universe: CORE.UNV
Comprised of dimension Tables ( D_TIME, D_PROD, D_STATUS, D_PROMOTION, etc… )
Derived Universe: PLANNED_EVENTS.UNV
Fact Tables: (F_SUMMARY, F_DETAIL )
Linked to CORE.UNV for the Dimension Records
Joins: All joins are done in the Derived Universe. You set up Contexts between the Two fact tables. Got to also define aggregate navigation and Aggregate Awareness with your measures(Which fact table to I pull my data from?)…
The above solution allows me to Navigate any of the Natural hierarchies which occur in the Dimensions Records (E.G. Year-Quarter-Month-Week- Day) and seamlessly pull data from any of the 60 plus measures in the corresponding fact table (Summary or Detail ).
Pitfalls:
*Data Modeling: Need to understand your source data and requirements, Build a Logical then Physical model. Define granularity… hire a pro…shameless plug 
*Data Extract Tool: Get an out of the box tool to help with the data extracting… E.G. ETI, Informatica. Its WELL worth it! 
*Universe Linking: Things can get all messed up when you compound the situation by working with multiple domains (DEV/TEST/PROD). E.G. you can have the fact tables pointing to TEST and the Dimensions records pointing to PROD. 
If you want some screenshots of our Universe Design let me know I would gladly share them with you.
-Dave

David Smith
(BOB member since 2002-07-10)