Near-real-time data warehousing defines how updates from data sources are combined and transformed for storage in a data warehouse as soon as the updates occur. Since these updates are not in warehouse format, they need to be transformed and a join operator is usually required to implement this transformation. A stream-based algorithm called X-HYBRIDJOIN (Extended Hybrid Join), with a favorable asymptotic runtime behavior, was previously proposed. However, X-HYBRIDJOIN does not tune its components under limited available memory resources and without assigning an optimal division of memory to each join component the performance of the algorithm can be suboptimal. This paper presents a variant of X-HYBRIDJOIN called Tuned X-HYBRIDJOIN. The paper shows that after proper tuning the algorithm performs significantly better than that of the previous X-HYBRIDJOIN, and also better as other join operators proposed for this application found in the literature. The tuning approach has been presented, based on measurement techniques and a revised cost model. The experimental results demonstrate the superior performance of Tuned X-HYBRIDJOIN.