That is exactly what I would expect. The fact that one text editor or display tool shows the text the way you expect, and vi shows it some other way, doesn't alter the data itself in any way. It's the same sequence of bytes in each case, but the non-ASCII bytes are just being interpreted in two different ways.

Check your docs for vi to see whether it offers any method for treating data as utf8-encoded. If it doesn't, you'll just need to accept the fact that "wide" (multi-byte) utf8 characters in your text file will show up as multiple single-byte characters (in the non-ASCII 0x80-0xff range) when you look at them in vi.