I tested this function, and did not find any discrepancy. Probably the following information could help you to solve your problem.
Rows of data from "Multiple Voxels Extraction" is stacked in a way that "subjects in conditions in group". i.e. it should look like this:
group1 condition1 subject1
group1 condition1 subject2
... ... ...
group1 condition2 subject1
... ... ...
group2 condition1 subject1
This is in the same order as the datamat. Thus, it is very convenient for Seed PLS.
On the other hand, rows of data exported from "Response Function Plot" is indicated by "st_evt_list" variable, and the different numbers in this variable represents different conditions.
If problem persists, please upload all the .mat files to your ftp server, and send me the link, so I can duplicate your problem.
Jimmy