From the report, just before the graph -- "The net load is calculated by taking the forecasted load and subtracting the forecasted electricity production from variable generation resources, wind and solar"

In other words, after projected solar generation is accounted for, the peak remaining power need for the specific grid in question is at 6-9 pm. (Other parts of the paper describe the "belly of the duck" being caused by "supply from solar generation resources" -- reducing the need for the rest of the grid to supply daytime power.) The paper in question may also be describing a residential-only grid.

Residential peak usage is from 6-9 pm, but when commercial/industrial power usage is accounted for, the peak is more like 2-6 pm. See the chart at around the 2/3 point of http://www.mpoweruk.com/electricity_demand.htm . Also note that residential demand may peak earlier on super hot / high sun days, as shown at http://blog.opower.com/2012/09/hot-and-heavy-energy-usage-ho... , and tends to peak later in the winter due to electric heaters kicking on in the evenings (https://www.gov.uk/government/uploads/system/uploads/attachm... ).

Of course, I'm glossing over a lot of complexity here. Each region is going to be different depending on a lot of factors. But in general, having a decent amount of solar power generation can help meet demand at the peak of commercial/industrial demand.