I suppose you could concatenate the two data sets, being careful to build the data so that each item has its own unique column (so if you have 12 items in the first MaxDiff and 21 in the second that your new design has 32 columns (one dropped out for the reference item). This should run but the utility scales you get for the first and second MaxDiff may not be comparable because you have no connectivity. While this seems like a possibility I am not sure how well it will work for you.
Another way to do this would be to run (e.g.) 2-12 class solutions for each model using LC and then take the segment assignments into CCEA (Convergent Cluster Ensembles Analysis) treating this as a custom ensemble. You could even get fancy and let CCEA go to town building ensembles from various cluster analyses of your HB utilities from the MaxDiff exercises and combine those to your LC solutions in your custom ensemble.