Odd that the figure says that the efficiency was 44.7% +/- 3.1% -- given the big error bar, how could they say they'd increased 1% in 4 months? Measure it again, who knows what you'll get!
There aren't even error bars in the article's graphs
Care to elaborate?
The +/- 3.1% is relative to the efficiency. The way to read it is: 0.447 +/- 3.1%, where now it's more obvious that it translates to more like 0.447 +/- 0.014.
Even if it really means 0.447 +/- 0.014, an increase of 0.01 is not very significant!