calculate entire RMS of an audiofile ,are audio statistics accurate?

Giovannivolontè  wrote if at tollerance 1% , for audition clipping is -3.97db L -6,12dB i really i don't understand it

Technically that's a bug, as in your example, the duration is zero and you can't have a clip lasting zero time! I suspect this has a lot to do with the rapid rate of change at the start of your selection. You will also note that in Amplitude Statistics it says Possibly Clipped Samples - strictly you need to consider the waveform either side of any given sample in order to estimate whether it's part of a clipped sequence. If you make a selection within a file, the rules say that you can only consider the selected part, but that you can't exclude the possibility that in a rapidly changing selection at the start, that the first sample could be a part of a clip in the section immediately prior to your selection - so basically it's playing safe (although I don't think that really, it should do this). But as it's not the easiest thing to detect, there is allowance for errors, and they aren't all obvious. For instance, it's quite possible to create sequences of samples that will overload the output of your sound device without technically clipping at all - by anything up to about 8.5dB. And you need, especially if you're going to normalize a file, to take account of this. Now, these sample sequences are actually legal - even though the clip detector would say they were clipped. So, this is not an exact science, and you have to use a bit of common sense with it.

No, I am not expecting you to take my word for the overload situation. Years ago I created a document (using Audition 1) showing how this can happen, and since this is all based on the Laws of Physics, nothing has changed. You may have a look at the document and its waveforms here: Dropbox - Audition_1.0_overshoot_doc.pdf