The rule of thumb I've used is to maximise the volume at the source, and manually set the volume using the analogue control (i.e., option 1). The reason is that the digital/software representation of the audio ought to be mostly distortion-free (no clipping, etc.). There will be quantisation noise in the digital signal, so increasing the amplitude may improve the signal-to-noise ratio (assuming the quantisation noise stays constant). Any distortion introduced by the analogue system may be amplified by the analogue gain circuitry - so it may be best to set the analogue volume as low as possible.
However, the disadvantages that may occur are:
- Distortion, if the software audio processing does clip the signal at high-volumes
- Eventually you may wear out the hardware analogue volume control
So I would experiment with options 1 and 3 to see which best suits your software and hardware configuration, and which is the most convenient to use.