according to a not so recent paper (http://www.sciencedirect.com/science/article/pii/S0167947303003049), it is a good idea to Varimax-rotate the factors that have emerged by Partial Least Squares.
Given the concerns that rotation causes in Principal Component Analysis (e.g. components not orthogonal any more, loss of maximum variance), I'd like to know why rotation seems to work here. More concretely: I am confused because PLS embarks from a principal component type problem, but the rotation takes finally place in space of latent factors and not in that of the original variables (which is, as far as I got, it the source of all rotation trouble in PCA).
There is some mathematical formulation in the paper, which probably solves the question, but is not understandable for me (referring in particular to sec. 3). Mabye someone could provide some intuition.