This is how you do science: you repeat the method to test the hypothesis. The article hints at what these adjustments entail:
"These included the volcanic eruption of Mt. Pinatubo in 1991, which spewed sunlight-blocking particles into the atmosphere, as well as the collapse of industry in the Soviet Union or the economic growth of China, ..."
This is similar to Hansen's 1987 (iirc) papers, which were based on a random prediction of a volcanic eruption in a particular year but it turned out to guess the year wrong. Predicting such events, which have a short term effect on the climate, is a guessing game. The numbers were pretty close, but if you repeat the method and replace the projections of CO2 emissions and aerosol emissions from volcanic and other sources, then they end up spot on.
These days, with more computing power available to run more detailed models more times, they do many model runs with a the random natural factors, and end up with a spectrum of results. This allows confidence intervals to be achieved. Hansen, in 1987, didn't have the resources for that; just like Sverre Arrhenius certainly couldn't do that when he estimated a 2C climate sensitivity from his manual model runs in ~1897.