I think your on to something here!!!
Most of the photo cells that we in the trade would install [like the Intermatic "Ice Cube" ones that screw into a bell box] either use a PhotoDiode or a Photoresistor for the light sensing device. This would be the element directly behind the clear lens.
On these, manufacturer specs say to aim it to the North - in order for proper light sensing.
Even on these types, I have found many failures when they are directed towards the Sun [directed West]. On these, the lens is sunbleached.
Almost every troubleshot photocell has been directed towards the West. The few that weren't had been damaged by some form of physical abuse [something or someone fell on it
They are rather simple devices and easy to understand. The Ambient light detection is done with either a PhotoDiode or a PhotoResistor. With these guys, having X amount of light falling across them will allow current to flow [in the PDiode], or will have a lower Resistance and higher current flow [in the PResistor]. Both would work on a small 1 pole relay, in which the line voltage control circuit would be connected across NC contacts.
Time delay - in the form of an RC circuit [Resistor-Capacitor] is used to eliminate false triggering.
I would imagine [this is just a guess], that in Bill's situation, the light detection element was a simple PhotoResistor. Along with that, the lens was distorted from the UV light being directly applied to it at high levels, instead of being dulled from reflections. This, plus the possibility of the PResistor being kind of reduced from the high levels of UV light, would appear to have made the photocell only responsive to high and direct light levels.
This is just my $.02 - feel free to nuke it if needed