Your observations are similar to mine in that some cloud can actually increase panel output. I put this down to cooler panels being more efficient, and this is what got me into a "cloudy day" controller.
My power usage is low in a domestic sense, lights, my biggest useage, and my electronic workshop.
The other workshop, where I do light engineering and car/motorbike restoration work runs seperately off the Petter, max 5KVa.
Cooking/water heating via gas/oil and wood burner. I have an old fashioned "safe" to keep food cool, refrigeration not really needed with vege garden and meat on the hoof so to speak. This way of living was common a century ago, and for centuries before that. I see no harm in forgoing coffee makers, heated towel rails and all the other crap you city dwellers deem indispensable....
Would a Lister actually feel comfortable running a heated towel rail?
Clip a high impedance multimeter to one of your solar panels , and monitor the output voltage under varying light conditions. Low light levels effect current dramatically, but the voltage remains fairly constant and when in darkness it will suddenly drop. No commercial controller is designed to work at these levels.
The idea here is to create what effectively is an impedance converter, where the milliohm battery impedance is matched to the high impedance of a light starved solar array. The capacitor passively does this by charging at the highest possible rate, the peak stored energy then dumped into the bank as a series of high current pulses.This is the electrical equivalent of a ram pump, the battery bank being the equivalent of an air over water collector/pressure tank.
Now, connect a largish capacitor across the same panel, and connect a small load, a flashlight bulb will do. You will get a brief flash as the cap gives up the stored energy. This is the rudimentary concept of current collection over time.
On cloudy days Its better for me to get 16 hours of 22 amps, than 16 hours of 13 amps, the tradeoff is to take a peak current hit mid summer, but then nights are shorter and less lightinng time required.
Im not sure what differences there are in different brand panel performances in low light, I have a very old mitsubishi 80 watt panel, and 4 similar no name 100 watt Chinese panels that are new, these are what Im working with. As I said, its an experimental setup right now,, lacking any refinements such as over voltage protection for the bank.... purely a proof of concept . The next phase will be to add a voltage up converter to lessen the static battery bank voltage threshold, the equivalent of seriesing more panels.... here it gets tricky, as circuit complexity increases, so too does phantom current draw.
Right now we have a partial moon surrounded by scuddy clouds, Im seeing around 4 pulses a second, with a 4700mfd cap this is around 2 mJ, which is bugger all..... a few hundred mA at most.
A dull overcast day, this rapidly increases to 120/200 J, there is a nonlinear relationship between photons and current (Coulomb',s law)
In bright sun, Im getting around 250/300 J, around 1/2 - 2/3rds of the max, at this point the panels are effectively tied to the battery voltage with a 80/20 pulse ratio.... the cap cannot charge instantaneously.... hence the loss. measurement made more difficult as the cap dis/charge rate is logarithmic, so ball park figures here.
With solar, there are so many variables.
But, as you are on the natiional grid, non of this has any relevance to you.