Eric Schulte writes:
> This does look like a bug to me.  Can you isolate "where" the value of
> data in your example is first assigned the wrong value?

The error lies in how ob-sh tries to determine if it needs to process a
table or a vector and then forgets to check for 'hline, thus ending up
in the vector branch with the table data.

> Thanks for reporting.  I will save your example for when I next have
> development time.

Here's a suggestion for a quick fix.  I've also implemented :hlines
processing for ob-sh (it was ignoring that header arg unconditionally)
and added an experimental header arg :hline-string to control how hlines
get actually exported.  BTW, trailing blank lines in the output get
swallowed when re-importing to Org, I'm not sure if that is supposed to
happen or not.