The validation rules were clarified in issue #306 but we failed to indicate that packageId should be included in the list of id instances that should be both unique and checked when resolving references. This seems to be affecting the validation in the R emld package. @cboettig let's discuss the issues I am seeing.

In addition, we should be clear in the validation rules that we are ignoring the system attribute, so probably need to remove or edit this rule:

• When references is used, the system attribute MUST have the same value in both the target and source elements, or it must be absent in both. Frequently it is absent in both.