Hi Ron,
that character is not part of 7 bit ASCII, thus it is encoded differently into byte values based on the character encoding used.
On my Unicode enabled system using the en_US.UTF-8 locale, § is implemented as Unicode character U+00A7, encoded as the single byte 0xA7 in ISO-8859-1 (latin1) encoding, the two bytes 0xC2 and 0xA7 in UTF-8 encoding and the four bytes 0xFE, 0xFF, 0x00, and 0xA7 in UTF-16 encoding.
The RADIUS shared secret is based on byte values, which could differ if different encodings are used on NAS and NAC.
To avoid such problems, I would suggest restricting passwords to use 7 bit ASCII characters only.
Erik