You say a p-value of > than .05 is good and means that the factor is significantly different from 0.
However, in regression (OLS), it is always the variable with a p-value < .05 that is significant.
Also, this site [https://stat.ethz.ch/R-manual/R-devel/library/stats/html/shapiro.test.html] says that for this test a p-value should be at least smaller (< than 0.1
So, unless I am completely mistaken and the opposite is true the information on this site is incorrect.