Output filenames are currently generated from Resource.getFileName() which returns the filename and no path information for the file system resources used as input files. Clearly this is the wrong strategy as your usecase highlights. I'm in the process of fixing this now so that additional path components will be used until all input files are uniquely identified. We could just always use the full path for the input file, but I don't like that idea too much.
Autowiring is tricky. BeanDoc works in isolation of the Spring container and can document context files without having to instantiate the container and load all the dependencies and so on. This means it's easy to document any context file even if it contains EJB/JNDI references and so on. Autowiring is container-provided magic and I don't see a way for BeanDoc to provide this unless a BeanFactory can be realised without requiring the compiled classes and associated dependencies. To be honest, I'm not a fan of autowiring anyway and never use it as I think it has the potential for maintenance headaches in evolving code.
However, if you can think of a way to do it I'd be delighted to implement it for you
Public Key: 0xE855B3EA