Are X1 and X2 sample sets or variable sets? I've assumed the latter based on the phrasing of the question but it isn't clear.
If they are sample sets then it is possible, but the result may suffer if there is any bias between the two samples sets.
If they are variable sets then you need to have the samples overlapping as otherwise you cannot map the pattern of variation in each variable from X1 to the pattern of variation in X2. You can only look overall trends in the absence of overlap.