Sorry for the confusion,  I should have paid more attention to the situation. (English is not my first language, so although I do check the things that I have written. I tend to read what I think, I have written.)

From another database we get (with an ODBC connection) tables and thereafter the mutations on the tables into a SQLserver database. The source database is not exactly relational. The mutations from the source get processed in the tables (in SQLserver). So at that point we have the tables and the mutations (in tables).

The tables are not totally 'logical' and do not really fit into the 'model'. So we are using views to create a beter model of the data.
Persisting the views is easy, just create a tables of the views. But now we want the mutations on the view. So the Inserts/Updates and Deletes needed to have the persisted table to contain the same data as the view. (The updates can also be done with only inserts/deletes).

As usual this is a simplified representation, everything is done within economical constraints and the usual 'management' constraints, it has to be good/cheap/quick. 

Offcourse; I told the management that we can do good/cheap/quick, but only two of them at the same time. For me this is a challenging puzzle, but I do like that. But do not mind scripts/documents/advises/tips to get a better/faster/cheaper 🙂 result.
Ben