At the moment, section 7.22 states
"When the Value of the SourceNode is an array then the BrowseName of the TargetNode shall
be <V[N]> where ‘V’ is the BrowseName of the Parent Node and ‘N’ shall be the array index
number of the DataType of the Structure."

The confusion comes from the "DataType of the Structure" which may imply that the DataType have to be a structured DataType.
For arrays, this restriction is not needed, the approach also works with built-in DataTypes. I also do not read out of the text that it has to be a structure, but I just had a discussion with someone who had the interpretation.
Simple fix could be to remove the "of the DataType of the Structure" from the sentence above.

Another question, that would be nice to have at least a recomendation in the spec is how to express this in the type system. For structures I think it is obvious that you have those as subvariables, either optional or mandatory. For dynamic arrays, I would use the OptionalPlaceholder modelling rule and the HasStructuredComponent ReferenceType probably using < BrowseNameParentNode>[0] to not violating the browse name rule. But that is not that obvious, so having this mentioned in the spec may be helpful.