
When importing data from Excel, rather than limiting the data to be imported by means of an import specification, my normal approach would be to import all of the data, and then manipulate it in Access to produce a set of correctly normalized related tables as the operational database in Access.
Having imported the data into a single non-normalized 'master table', this is then decomposed into normalized tables by executing a set of INSERT INTO statements ('append' queries) in Access. In effect this creates an import specification in the form of the set of 'append' queries, so the process can then be repeated to import further data subsequently simply by executing the 'append' queries each time.
You'll find an illustration of this with a simple Excel worksheet in DecomposerDemo.zip in my Dropbox public databases folder at:
Extract the DecomposerDemo.accdb and ContactList.xlsx files from the zip archive, and in the former browse to the latter as the source for the import.
The basic ground rule when decomposing the imported data is that data is first inserted into the referenced table in each relationship before data is inserted into the referencing table(s). The first stage, therefore is to insert data into each table which models an entity type where the data can be inserted with a query which selects distinct values from one or more columns in the 'master table' alone, giving each table an autonumber primary key. Having done this, data is then inserted into tables which model relationship types between these tables. This involves joining the referenced tables first created to the 'master table'. My demo steps through this process, with a brief description of each stage as you proceed through the demo.
The demo also illustrates an interface with the imported data by means of a form/subform, in which there is also a button to open a simple report.
Although the demo uses a very simple worksheet as its source of data, the principles behind the operations in the demo are the same with more complex models. The key to a successful import is in recognizing what data can be inserted at each stage without violating referential integrity. This requires a reasonable knowledge of normalization, and a familiarity with building queries based on a number of related tables, but anyone designing a relational databases beyond the most trivial models will need these relatively modest skills in any case.