About the relationships graph
When you work with tables in the relationships graph, you are using them to organize your view into your data. Each table occurrence in the relationships graph represents a separate view into your data. When you join two tables, you are leveraging the two existing views to create a third way of viewing your data. For example, if you have an Invoices table with invoice ID and customer information, and a LineItems table storing product orders for each line of each invoice, you must create a relationship between the two tables before you can display data from the LineItems table on the Invoices layout.
You can create a relationship between any two tables in the relationships graph, but the relationship must not create a cycle, or closed loop between tables. That is, each series of relationships must have a starting table and an ending table, and those tables must be different tables.
Because each relationship represents an additional set of criteria, you must be aware of your context within the relationships graph. Context is the point in the graph from which a relationship is evaluated. Because the graph is never a cycle, each point along the graph offers a different perspective into your data.
Related topics 
About relationships
About planning a database
About match fields for relationships
Creating relationships