The JSON-LD writer converts language tags to lower-case, e.g. "en-US"
becomes "en-us".
The code is in JsonLdApi:L664 and I guess it follows the recommendation
https://www.w3.org/TR/json-ld-api/#h-note-17
And the RDF 1.1 spec itself states that the value space of language tags
is lower case.
But syntactically that is a MAY and it means that RDF graphs cannot be
correctly round-tripped via JSON-LD.
Could there be a flag to switch this off? The other serializations also
do preserve the casing AFAIK, and when you add a triple to a Jena Graph
also no such transformations are taking place.
Thanks
Holger
