If I'm reading that study correctly, those figures have an R² of around 0.23, which is quite low for explanatory power of wages as dependent on union membership (quite low as in the "none" or "very low" range)