If the libraries were made using the same extraction from the same sample, the current approach can be used to obtain an overall estimation for that specific extraction of the sample.
If the libraries were made using different samplings/extractions/parts of the same individual, library complexities should be calculated separately for each library; which would currently require multiple superduper runs over each library.