If I understand right, that figure is measuring the accuracy of a prediction of electricity production, not the constancy of the production. We can't smelt aluminium with no power even if we have predicted that there'll be no power.

To be honest I don't know, but I do know taking advantage unpredictable power of this sort is why Aluminium plants are built in the first place. Where I live they are paired with coal fired plants. The reason they are paired with coal fired plants is coal can't change it's output fast enough to match the somewhat unpredictable changes in electricity usage.

So the coal plant needs someone who will buy power when it's cheap, and not use (much) when its expensive. Apparently an Aluminium smelter fits the bill nicely. I presume it uses enormous of amounts of power to disassociate the Al(OH)3 but needs only a small amount of power to keep the pots at operating temperature which is not particularly surprising as the pots can be thermally insulated.

Maybe the above claim that Aluminium can't tolerate rapid changes in available electricity is true, but it sounds odd. Coal fired plants often "trip out" - meaning the generator drops off line without warning. This is an unpredictable change that happens much faster than the sort of unreliability we see from renewables - it's a huge change that happens in milliseconds. And it's not uncommon: https://leadingedgeenergy.com.au/coal-fired-generators-trip/ Again, the Aluminium plants seem to cope with this extreme unreliability just fine.