Well, string.byte looks at the string as if it would be ascii, but in fact the nspire uses utf-8. UTF-8 uses multiple bytes, unlike ascii. And since string.byte is for ascii, it takes only the first byte, 194.
To make a story short, you need string.char(194, 178) for ². 194 is the first byte, 178 the second.
Now, to find the bytes of a utf-8 char first use string.len, to see how many bytes there are. Then use string.byte to fetch the chars:

Code: (Lua) [Select]

ch="²"
string.len(ch)  --returns 2
b1=ch:byte(1) --returns 194
b2=ch:byte(2) -- returns 178

chr = string.byte(194, 178) --returns ²