p values are only useful if the statistic is applied correctly...even then they are prone to misinterpretation
If you want to do a statistical analysis to see if something is significantly correlated you might not want to take the output of a robust method of identifying correlations (KEM analysis). You can be pretty sure they are significantly correlated. If you want to adjust the statistic for multiple measures it can be done but that is difficult, imprecise, and I expect AVXL doesn't do it or does it in an advantageous manner. It is no surprise they never spell out their statistical analysis in detail.