Architecture of SEM-BCS 4.0

I'm trying do understand how SEM-BCS 4.0 works and where the different processing-steps (mapping of R/3-fields to BW-InfoObjects, Validation of reported data via a rule set etc.) are done.

Is the reported data via a basic R/3-system transfered via the normal transfer-rules, psa and so on?

When is it useful to integrate a staging-cube for original data before the bcs-cube?

Do manuel entered data go directly into the bcs-cube or is there a step between?

The bcs-cube is a transactional cube. How can the problem of switching between manual entry and upload be solved if there is a demand for simultaneous upload and manual data entry?

I couldn't find any good information in the web about this! I would be grateful if somebody could help me with the basics of bcs.

Thanks in advance!

Greetings

Patrick