you will need to merge your datasets into one dataset.

RS doesn't support multiple datasets with a connection btw the datasets. However if you merge the data into one dataset and add one column with an enumerator(1 = first "dataset", 2= 2nd "dataset) you add additional tables with a filter set to the shared dataset pointing to that enumerated column be sure that you use the = as operator and =1 in the field for the filter. int data types require the =X instead of just X if it was a string enumeration.

bottom line is you can "sub report" without sub-reports using the shared dataset with a filter column and using the filter properties of a list or table. Be sure to populate the shared dataset with any data used to join the "multiple datasets" in appropriate columns that can be grouped on... such that each enumerated dataset will have its own data columns plus the enumerated column plus any columns needed for the grouping connections between the "multiple datasets"

I took a report with 26 subreports(based on 1 to N cardinality requirements) into a single dataset with a filter on each nested table. Prior to the change imagine 5000 records calling the databases one entity at time populating each sub report. After the change, not only does excel work with the nested filtered tables(ie virtual sub-reports) but sql only gets hit once instead of 5000x26 times. Let sql use the power of set logic and reporting services filter out the appropriate rows it needs for each nested table/list. Note that as you add additional enumerated sections the data spreads out diagonally down your table when observing the output in sql. While there maybe a lot of white space this is currently the only option short of using report builder off of a data model that which lets you connect multiple data sets together. It is very important that you try to grasp the single DS/filtered concept for 1 to N cardinality/multiple data set situations. This is a make or break concept for RS and it pains me and Teo Lachev to see that MSFT hasn't allowed us a more efficient way to resolve multiple data set requirements.