There is no single best way to do this type of transformations. There are many unexpected complexities that arise - hence the lack of support from the company as well (not to mention that there is no financial incentive).

I personally think that the proper way to deal with a situation like this is to merge the results of the analysis not the data itself. Think of if as a replicate analysis, can your old data with the old annotation produce the same or similar outcomes as the new data and new annotations.