Why a Feedback Diode?

[LoveLaika]

I'm looking at the design for the LT1765 switching regulator. I previously used it in an old project, but I wanted to make some improvements, specifically for the inverting-buck boost configuration. I was looking at the datasheet for it when something caught my eye. A diode (D3) in Figure 12 connects the feedback node to the negative output. Why is that there? If you look at the block diagram, it goes between the inputs of the error amplifier. The feedback node would be more positive than the GND terminal (which would be the negative output here), so that diode just ensures that the error amplifier difference is maintained? I don't see how the diode would ever conduct.

For what it's worth, I tried to use the configuration previously to convert 13.8 V to -10 V. It worked, but it could not handle a 1-amp load despite using a 10 uF inductor. It could only handle a 0.5-amp max load while maintaining the correct output voltage. According to my calculation sheet, a 10 uF inductor should be able to handle that amount of current. Incorrectly following the design shown in Figure 12, I had the cathode of D3 connected to the common ground instead of the feedback node, so I'm wondering if this would have made all the difference.

[David Hess]

If there are loads between -5 volts and a positive voltage, like +5 volts, diode D3 prevents those loads from pulling the output positive which could prevent the switching regulator from starting.

[LoveLaika]

Thanks. I understood that explanation.