It's been pointed out that the spec doesn't distinguish between supporting files contextualising the data and the actual research data parts of the package. Have a look at how to address this - how is this done in the archive/preservation community? Is there a way to model this in schema.org?