You are on the right track with thinking that grouping is the key to your best solution. My suggestion is to incorporate all the values from the upper tablix dataset into the same dataset you are using for the second tablix. That will result in a lot of redundant information in the second dataset results, but then if you group the tablix by a value they all share (like the primary key for the table that lies behind the first dataset), then put all that information in a header row of that group, it will print only one time for each iteration of that value.

With this solution you can delete the top tablix and just use one tablix grouped by whatever value the top tablix had to identify each row uniquely.