[charlie]

With surfaces which are not parallel, there will be some diffraction of the sound beam as it either enters or leaves the sample, just as in TTU. It depends how “unparallel” the faces are. “Stepped” surfaces should not really present problems however, as long as you consider the increase in thickness with more plies.
It also depends on the orientation of the sample. If it is not perpendicular to the transmit sensor and the screen, then sound will again be diffracted.

With respect to your question about structures other than thin composite panels, the operating frequency will be the determining factor. If I remember correctly, the Santec systems acoustography operates at around 3 MHz, so is not suited to very thick monolithic Carbon Fibre samples, nor to sandwich structures or Fibreglass.