The hold music played by the voicemail is heard only when the voicemail reaches a busy extension, then asks the caller if they want to hold and the caller chooses to hold. During this hold time, the vox file is played for the programmed time to the caller, then the voicemail tries to ring that extension again. In your situation, the MOH that is connected to the processor should be played during a supervised transfer. Check the MOH from the vm port.