There are 2 main issues with that approach
1) SQlite and MS Sql both conform to some quasi-sql-like-query-form, however, their metadata/administrative structures are radically different, getting table mappings and key restraints etc. and are handled differently in the end. As a result, since about 70% of the sqlmetal logic is just extracting that information about a database, there's really not much to be had.
2) This code is to be open for anyone to use, preferably open-source (although there was no license on the tutorial I followed for the first parts of it, so hypothetically someone could make a fuss, but it would probably die/ who posts a tutorial but presses copylaws on the code). However, this same generous nature most definatly does not apply to decompiling/reflecting the Microsoft libraries, they come down fast and hard, and there is a good chance if the code was ever incorperated into a larger project, that Microsoft migth try to shut down both of them.
=/ I know it kinda sucks, but the main issue is I rarley use SQlite outside of the most simple capacities, and as a result, I'm not familiar with how constraints etc. are handled (I think SQLite ignores alot of them..) but yeah, if somone who really knows their sqlite wants to help, all I need is a valid mapping tool, and I can finish a buggy implementation.