Really interesting post.
Does it have the same weakness as Enigma devices, i.e. that the input char can never become the output char ( encrypt(a) is never a)? This significantly reduces the search space.
Let me explain, the M-125 uses a card reader to alter the input characters sent to the rotors. The rotors aren't being applied to raw input.
input -> card transformation -> rotors/reflector/rotors -> output
>"The activated wire corresponding to the input character wire goes to the card reader (the slot on the left side). Holes punched in the card control the swapping of characters. So, our A is translated to a T. [then it is sent to the rotors]"
After the card reader transformation the "never the same character" property would hold, but the card reader makes it significantly more difficult to exploit. That is, if you could guess the output after the card reader transformation you can apply the attack, but you'd probably have to guess the card reader transformation key first
Note that the Enigma had a plugboard[0] for a similar purpose but was less effective since most characters were left unchanged by the plugboard.