Hmm... "why"-questions about implementation aspects typically are not satisfactory for anyone not involved in the actual implementation...
But I'll give it a shot anyway 
Ok, so what's the problem you solve with an analytic view?
It's that of the data "cube" based on a star schema.
The star schema (and its variations) is a modelling approach where you have all the transaction information - the "facts" - in a central fact table (great naming going on there).
The relevant descriptive information for these facts, aka master data are linked to this fact table and called "dimensions".
Choosing this name only makes sense once you realize how the "cube" is supposed to be used.
The idea here is that you pick the dimensions of your interest in the data and select specific values for each of the dimension you find interesting. With this choice of specific values for dimension you now look into the facts table and aggregate what ever data you find there matching your selection.
In terms of database tables it means, that you want to specify filters on the dimension tables (on the outer tables of the star schema) and by joining to the fact table these filters are applied to the fact table.
Whatever data of the fact table is not filtered away gets grouped and aggregated.
Notable here is that the grouping is based on dimension fields and the aggregation only happens on the so called "measurement" columns in the fact table.
For this funny construct, you don't want to employ classic table-by-table-joining. Instead it would be great to actually be able to combine the different filters from the "outer tables" at once to the fact table and perform the aggregation then.
THIS is what the analytic view does for you.
It is highly optimized to perform this very specific query pattern and lots of the optimizations are build upon the column store table internal data structure.
That's why you cannot have row store tables in an analytic view and that's why you can have only one fact table with measurements in it.
There you go - now you know.
- Lars