In this dataset, the Sales Territory, Product, Date, Currency, and Promotion dimension tables are related to both the Internet Sales and Reseller Sales fact tables. As shown in the following screenshot from the Relationships View, these relationships and the three dimension tables specific to either fact table all have single-direction cross-filtering enabled:
Unlike the shared dimensions, the Reseller and Employee dimension tables are exclusively related to the Reseller Sales fact table and the Customer dimension is exclusively related to the Internet Sales fact table. This is a common scenario for larger models in that fact tables will both share dimensions and maintain their own exclusive relationships to certain dimension tables.
In general, it's recommended to avoid bidirectional relationships between shared dimensions and fact tables when there are also dimension tables exclusive to certain fact tables. This is because such relationships generate filter contexts that business users often don't expect or desire and that don't add analytical value. For example, if the relationship between Promotion and Reseller Sales was revised to allow for bidirectional cross-filtering, a report that analyzed internet sales by customers would be impacted by the filter selections of the Reseller and Employee dimension tables even though these two tables are not related to Internet Sales.
A better use case for bidirectional relationships is between the exclusive dimension tables and their fact tables, such as from Reseller to Reseller Sales or from Customer to Internet Sales. These bidirectional relationships aren't required given the CROSSFILTER() function and other options available in DAX, but they allow simple measures against these dimensions, such as the count of resellers to reflect the filter selections applied to other Reseller Sales dimensions, such as Sales Territory and Product.