65% of the noise disappeared "like magic".
25% left some "amp too loud" traces, erratic frequencies, crescendo, not too loud, needs a second pass to reduce them below audible level.
Rest is these damn "100% hits". 5% you get, the rest is a decision of "what is worse?". In a cascade of "influenza" (you know, this "one starts and the waiting rest follows" situation) only one(#4) rests audible, if played very loud.
The thing is: you can see these noises in the frequency display. If they are separate from the main signal you can marquee them and do "magic". If they are spreaded, you have a chance if the disturbant signal is not too long so that you have enough sample information to make it possible to identify and separate the two "noises". So any narrow signal is removed 100%, like the chair+pedal noises.
A very powerful feature. Will Audition end up, with V. 3, like Photoshop, as THE editor? Hmm..., possible.