Thanks for pointing out the commentary, Andrew.  The issue they point out has to do with the interpretation of behavior PLS analysis and "activation" as I read it, rather than group differences, which comes from the task PLS.  The authors do make some legitimate points but are somewhat inaccurate on their own interpretation.  I will write a reply to their commentary after my grant goes in :)

To answer your question, a task PLS done with more than one group will test group differences and there are several examples of this publshed that you can refer to (the Diaconescu paper cited in the commentary is one).  The issue is whether these differences are best interpreted as "activation" differences.  I would assert that they are, but you do need to look at the nature of the changes to be clear on what the cause of the difference is.  Here looking at HRFs or group means is needed.  Remember that activations in neuroimaging are almost always relative measures so you need to look at the raw data to know what is changing.  Its worth noting that this requirement is there for multivariate and univariate analyses alike.