I dont know what more to say but no matter what length of audio is highlighted, the beginning 5 to 20 ms (rough guess) is attenuated much greater than the remainder of the audio. It is vocal also but probably does not matter. I just wondered if at all possible there is someone who works at audacity or related that could tell me if this is a glitch that can be fixed or somehow intentional (cant see why). Uncheck compress based on peaks and make up gain boxes which from what ive read makes this compressor into rms mode. I just think the algorithm is not right.