Difference between CHAR(n) FOR BIT DATA and BINARY(n)?

I think I have seen somewhere that there is some sort of semantic difference between CHAR(n) FOR BIT DATA and BINARY(n), but I'm unable to locate where it was. Any pointers? I'm asking because I have loads of columns defined as CHAR(16) FOR BIT DATA from before BINARY(16) existed, and thinking about gradually migrate them to BINARY(16). I'm curious whether there are any circomstances where migrating one side of a f.k. relationship before the other will cause problems.