Hi Johan,
You understood correctly. The former implementation was not ideal because sounds coming directly from the back were completely silent.
For front-back differentiation in headphones, the preferred solution is to use a binaural decoder such as Google Resonance or Auro HP. Independently, whether you use a binaural decoder or rely on built-in decoding to stereo, you may use the Listener Cone built-in game parameter and map to volume in any way you like. You can even replicate the former behavior.
Hope this helps