@yaxu first question after reading through the thread and looking at your samples as reference:
Automatically topped and tailed and normalised to -3db with ecasound/sox
I've normalised mine to -1.0 db... is this something we should be making consistent from the outset?