This is not exactly correct. The link explains why by breaking down the stages of efficiency loss. The solar panels only mitigate the first step, so you are look at 22% vs 47% for plants. Then, the 22% is not including losses from converting the electricity to lighting (or heck, the energy to create those panels in the first place).
LEDs can select the most favorable wavelengths however so maybe it can approach parity again.
You could generate chemically costly preproducts for the plant synthetically too to help its efficiency.
Things like increasing CO2 and providing at least nitrogen.
But while you are at it, you could directly to synthetic food... Hydrogen, methane, formaldehyde, sugars.
I wonder if it's really necessary to convert the light to electricity as an intermediate step. I'm imagining a filter that's mostly transparent but has a fluorescent coating to convert UV to visible light that the plants can use, effectively brightening the sun in the useful part of the spectrum. Unfortunately Google is giving me the wrong kind of results for "fluorescent filter".
The material you are looking for is called a scintillator. It converts higher energy photons to visible. Of course it has its own issues / cost. Plus, you have to get the light from the rooftop to plants.
1-2% of best plants vs 22% of best commercial panels! :)