There are situations where this module may be truly useful. E.g., I've had an electric fence controller running off of a 12V lead-acid battery and a PV panel of about 4 watts with a simple anti-overcharge controller (VW dealer type). It barely kept up - sometimes after cloudy weather I had to bring the battery elsewhere to charge it. Switching to this module might have reduced the number of times I'd had to do that, if the resulting PV charge rate was noticably higher.
Going back to the question of can the _input_ voltage be regulated via the feedback pin, without losing the output regulation, perhaps yes after all. That is because, when charging a battery, we don't really want to "regulate" the output voltage, we just want to limit it, so the battery will not be overcharged.
Think of the following circuit: same as OperaHouse's above, but connect the collector of the NPN transistor not to pin 5, but to pin 4 via a resistor. Arrange the resistors all around so that when that transistor is conducting, Vout will be equal to the highest voltage we want to allow the battery. If the PV voltage is high, and the battery is full, the transistor will be "on" and the LM2596 will see its target voltage and not output any current. If the battery voltage is less than full, the LM2596 will increase the output current in an attempt to make the output voltage higher. That will eventually lower the PV voltage and turn the transistor off. When off, the feedback pin sees a higher voltage, and the LM2596 will stop the output current, causing the PV voltage to increase again, etc. Ideally the transistor will settle somewhere between all-on and all-off. That can probably be arranged by adding a small resistor between the emitter and ground, along with a resistor between the base and ground. Add a filter capacitor or two and no more "wham bam". But will that be more _efficient_ than using pin 5?