On IBM i, CCSID support is intended to be seamless. Imagine the situation where the table is in German encoding, your job is in English and you are creating a new table in French - all on a system whose default encoding is Chinese. Use the appropriate CCSID for each of these and the operating system will do the character encoding conversion for you.
Unfortunately, many midrange systems aren't configured properly. Their system default CCSID is 'no CCSID / binary' - a remnant of a time some 20 years ago, before CCSID support. DSPSYSVAL QCCSID will tell you what the default CCSID is for your system. If it's 65535, that's 'binary'. This causes no end of problems, because the operating system can't figure out what the true character encoding is. Because CCSID(65535) was set for many years, almost all the tables on the system have this encoding. All the jobs on the system run under this encoding. When everything on the system is 65535, then the OS doesn't need to do any character conversion, and all seems well.
Then, someone needs multi-byte characters. It might be an Asian language, or as in your case, Unicode. If the system as a whole is 'binary / no conversion' it can be very frustrating because, essentially, the system admins have lied to the operating system with respect to the character encoding that is effect for the database and jobs.
I'm guessing that you are dealing with a CCSID(65535) environment. I think you are going to have to request some changes. At the very least, create a new / work table using an appropriate CCSID like EBCDIC US English (37). Use a system utility like CPYF to populate this table. Now try to download that, using a CCSID of say, 13488. If that does what you need, then perhaps all you need is an intermediate table to pass your data through.
Ultimately, the right solution is a proper CCSID configuration. Have the admins set the QCCSID system value and consider changing the encoding on the existing tables. After that, the system will handle multiple encodings seamlessly, as intended.