I think the problem is the double usage of the UTF-8.
For me the following worked:
- pasting the plantuml code straight into the plantuml webbrowser (see http://www.plantuml.com/plantuml/png/SoWkIImgISaluKh9J2zABCXGI4nKgEPIWB2ZapCIYufJIsoLGaFST2sKOgsv5F8C5Q9p4fDISsoL7ZUtFMxQ-RYvGqb78Iq_hoGlCJVLLiatMC7TFyJ3qrafbgkMoo4rBmLeF000)
-same as above, even without the graph... line (see http://www.plantuml.com/plantuml/png/SoWkIImgISaluKh9J2zABCXGI4nKgEPIo3DKYCv9J4hDiLLwtDpsk6ddu-OE9Hs5jFoyahp4t5HR9TzW1NV_40zFPwLOhbekXzIy5A0X0000)
Note that I'm not on a Japanese type of computer / OS.
Running with the newest beta version (PlantUML version 1.2018.14beta12), I got the following results:
- code as specified in question
- call with -charset UTF-8: resulted in question marks in picture
- call without -charset UTF-8: resulted in a good picture (as far as I can judge)
- code without the graph [charset="UTF-8"]; line:
- call with -charset UTF-8: resulted in question marks in picture
- call without -charset UTF-8: resulted in a good picture (as far as I can judge)