I feel your pain. I haven't done anything with JSON in SQL Server but I know that JSON Path, unlike XPATH does not have a getParent() equivalent function which means that bulk ingestion of JSON containing arrays produces two unconnectable recordsets.
The approach we've had to address it is import a file containing many JSON documents of the same document type and loop over each document triggering multiple extractions.  In effect reinventing RBAR.
I am currently experimenting with ways of bulk ingesting the non-array part of the document, which is very fast, and then submitting those documents containing arrays to the RBAR process.

I've had some luck with a library called YAJL but no son of mine will every be called JASON.