The possible point for using digital zoom in, say, 1200x1600 mode instead of the full 2592x1944 of the sensor is in the situation where the digital zoom is cropping to at most 1200x1600 of the sensor's pixels anyway. Then outputting at 1200x1600 simply avoids useless low quality upsampling of that lower resolution crop, whic takes up extra card space while adding no information. Avoding that in-camera upsampling (a) saves camera storage space and possibly card writing time, and ( leaves you with a better (less processed) starting point for any higher quality upsampling that you might want to do on the computer.
P. S. On the other hand, it seems to me that in general, when you need to get down to a given file size (say 500KB, maybe to fit the limits of emailing or web site hosting), it should be better to do this by greater JPEG compression than by downsampling to a lower pixel count followed by using less JPEG compression, still with the same target file size. E.g. 8x compression versus halving pixel count and then 4x compression.
My thinking is that both methods are discarding the same amount of data to hit that 500KB target, but JPEG compression is designed to make far smarter choices as to which data to keep and which to discard than crude downsampling.
