First of all, I have to say this module is a gem.
You have to get used to it but then it works like a charm. I have CSV tables with 10k+ entries, several cross references and multi-value fields. I've tried other import modules. This one is working.
Thanks to the Developers!

I have a small issue with node references. In my setup upper and lower case node titles refer to different entities. Node import, however, always chooses the lower case version. This can be seen as a collation issue in the mysql, but users are ofter unaware of this and might also run into this problem.

I think title lookups in node reference fields should have a configurable option to use the BINARY operator in string comparisons. Maybe in step 5 among the additional options.

The function node_import_node_reference_potential_references() could be extended in a way that SELECT statements include the BINARY operator when required.

Btw. I'm using the patch from #514552: Importing Node Reference fields relies on unique node titles (not always practical).