You could try
creating a new speech recognition profile, so the training data from your previous microphone does not affect the recognition on this one.
Also make sure to check the "Disable Adaptive Recognition (useful for noisy environments)" option on the "Recognition" tab of the VoiceAttack options window.
Though, you're combining what by necessity is a microphone that will pick up loads of environmental noise, with very short commands, which tend to be difficult to recognize correctly and consistently even with a good microphone.
Voice recognition (not just VoiceAttack) works best with a microphone that picks up your voice very clearly, without picking up background noise, and command phrases made up out of a decent number of distinct phonemes.