Clarify if the LDD structure is for storing the model ONLY or to also create\author LDD in Collibra?

I think it is misleading to assume there was use-case based design decisions.
There is no “Logical data dictionary” functionality, there is just metadata management functionality, and it so happens that in the list of asset types, you can also define assets within a LDD.

Therefore, it is up to the users to define how to use those asset types and the processes to collect the metadata. You need to define for each activity what makes the most sense for your organization (creating a model, reading a model, relating a model to other metadata, etc.)

In my opinion, the modeling capabilities are very limited:

• It is difficult to create a model: lack of most of data modeling capabilities, including semantics such as foreign keys, constraints, etc.
• Difficult to read a model (the business diagram feature does not allow for layouts, custom colors, notes, etc. that are usually needed for data models
• But it is great to relate to other concepts and support data governance activities: Guided stewardship, etc.

In that context, I do believe Collibra is intended for “storing” only, for the sole purpose of supporting governance processes. Modeling and consultation should mainly happen within a data modeling tool (or format supporting similar rich visual capabilities).
The fact that you CAN edit the data in place does not mean you should configure it like so. You could keep the metadata in a domain where no regular user have any edit capabilities. This is similar to the physical data dictionary: they are not meant to be updated in Collibra, they are supposed to be registered from connectors such as Edge, marketplace connectors, etc.
The most you could hope for is “enrichment” of the collected metadata.

Thanks, @arthur.burkhardt, you have resolved a big concern.