That sort does not behave as intended. The -k5.1,5.1 restricted key never has any effect. While it restricts that key to the first character of the fifth field, since leading blanks aren't ignored by default, when using the default delimiter, 1-index is always going to be a whitespace character.

Code:

$ cat data
2 2 2 2 a 2
1 1 1 1 b 1

# 5.1 for both lines is a space, whole-line lexicographical comparison breaks tie
$ sort -k 5.1,5.1 data
1 1 1 1 b 1
2 2 2 2 a 2

# -b ignores leading blanks so that 5.1 yields 'a' and 'b', correct result.
$ sort -b -k 5.1,5.1 data 
2 2 2 2 a 2
1 1 1 1 b 1

Note that if the b modifier is used in a key definition, the modifier only applies to the component to which it is attached, and the -b option (if present) is ignored. Therefore it's possible that even within the same key definition, identical indices may refer to different positions. Given the three character field, ab (space character followed by two letters):
5.1b,5.1b ==> a (1 character)
5.1,5.1b ==> a (2 characters: space, a)
5.1b,5.1 ==> (0 characters: null result since the second component points to a character preceding the starting component)

Regards,
Alister

EDIT: The thread moved on while I was preparing this post. Perhaps it's no longer relevant to the solution, but the information may still be of use to someone.