it seems that the normalisation preferences (from the Audio preferences) do not really influence the behaviour of the sound synthesis. This is particularly relevant when synthesising a list coming i.e. from a loop.
Meaning: at the end it will be normalised to 0.0dB, whatever the Audio preference is (even if left unchecked), and only in the case that the peak goes over 0.0dB.
It is not a dramatic issue, since in every test I performed I got “0 samples out of range” from the listener. Adding the keyword “rescale” to the synthesize object works, normalising to the given dB value.