Modify the DTS package to check if it exists or modify the job to do the check. It's probably easier to just modify the job. Before the call to dtsrun add an IF statement that checks for the table(s). Only run the package if they exist.
You're going to need to write a bunch of code. I think I'd find out what tables were there via INFORMATION_SCHEMA.TABLES and then bcp those out to files. Once the data is in the files, I'd have another process that looped through a folder and processed each found file. Perhaps you could name your files the same as the table names to make the import dynamic and easy.