Detailed help for this control is almost impossible to get. If it wasn't for the source code being there it would be just about useless.
I had to sift through the sources to get a better understanding of what is happening here.
When using the "SaveTostream", it writes a magicID at the start, something that is standard in many things like Image formats.
The details for the nodes are written but if you have custom data that is sensitive to memory locations you can't be using any managed types in the Node record, which is an attachment to the end of the Record. What you can do is create your attachment, basically added fields that would indicate an Offset and size of your data so that it can be reloaded at a later date. No using absolute addresses unless this is data to be reloaded with the same session.
The OnNodeSave event gets called after the Node record is written and that is when you write your custom data outwards, assuming the details of the info are properly up to date in node record because at this point that has already been written.
etc.
Hope this helps anyone else, I am still sifting through it.