At the moment when you are using upper case subfields in items and your bibliographic records, this will work in the frameworks, in cataloguing and in searching with Zebra. But as the item search uses SQL and ExtractValue, it will always return the contents of the lower and upper case subfield, no matter what you configured for your additional item search fields. Example: 520$a, 520$A
Sadly I am not sure we will be able to fix this. The biblio_metadata has a case insensitive collation right now. Which I think is what we want when searching for example for a title or author in the record. I am not sure we can treat the attributes (MARC fields and subfields) in the MARCXML record as case sensitive, while treating the contents of the elements as case insensitive at the same time.