Based on the behavior you described, this looks like a Managed SharePoint ingestion bug or preview limitation, not an obvious error in your configuration.

Databricks’ Excel reader supports selecting one worksheet with dataAddress, and your successful read_files() test confirms that both the worksheet name and Excel structure are valid. Databricks also documents that the Excel parser reads one sheet at a time and uses dataAddress to select a different sheet.

The suspicious part is this combination:

• Both ingestion objects use the same SharePoint file URL.
• Each object specifies a different dataAddress.
• The correct row count is read from each