A common way for exposing RDF data on the Web is by means of SPARQL endpoints which allow end users and applications to query just the RDF data they want. However, servers hosting SPARQL endpoints often restrict access to the data by limiting the amount of results returned per query or the amount of queries per time that a client may issue. As this may affect query completeness when using SPARQL1.1’s federated query extension, we analysed different strategies to implement federated queries with the goal to circumvent endpoint limits. We show that some seemingly intuitive methods for decomposing federated queries provide unsound results in the general case, and provide fixes or discuss under which restrictions these recipes are still applicable. Finally, we evaluate the proposed strategies for checking their feasibility in practice.