Look at the graph in figure 2 - the massive overlap in the interquartlie ranges suggests that any differene in the median values can be attributed to random var

Look at the graph in figure 2 - the massive overlap in the interquartlie ranges suggests that any differene in the median values can be attributed to random variation within the samples, rather than a true difference in the populations. Also its better to quote an absolute rather than relative reduction, so the difference is really 3.6%