Pauses are one reason. Another reason can be the mic position. If you mic is sitting directly in from of your lips, it can register berating as well and that can cause the system to miss the words you are speaking. I was advised about this issue before taking the test and fortunately was able to score 90 each and 87 for speaking in first attempt. Funny thing is that in speaking, during repeat sentence, I was able to repeat only 4 sentences completely, rest were partially correct (but I didn't umm or aaa, i just kept speaking)