It is entirely possible to do. I manage an authoring environment developed specifically for this purpose.
However, you will probably have to specialize your DITA implementation* (or go totally off-road and brutally hack your DITA implementation or just use a custom doctype). Regardless, you will need to customize your (or develop a) stylesheet to manage the suppression and aggregation. Because you are going to be using specific elements (or, possibly, attributes), probably for each column, your authors/customers will have to agree to less freedom to alter table contents and format (and what freedom you have will come with project lead time to write requirements, develop, QA, and release changes). I suppose there is a path based on code that generalizes the suppression and aggregation, but my gut reaction is that that is a lot more code than supporting explicitly enumerated table types, column names, attributes and values, etc. (Maybe I do not understand the rigidity of your authors' table requirements and it is not as complicated as I imagine. My experience has been that there are always changes, initially unstated limitations, emerging requirements, and the normal sorts of scope creep leading to more complexity than first appears.)
In short, this is not a path to take lightly without a lot of careful thought and exploration.
* I assume you are in DITA based on your use of the terms topic and bookmap. If you're not in DITA, little changes regarding what you must do, however, you may lose less as you are not giving up DITA compliance and the support that brings.
