Is it possible to generate csv file with column labels of child records even when they don't exist?

Summary:

Standard BI Publisher behavior is that column labels are not generated for columns of child records when they don't exist.

Example:

Data Model queries:

Data Set Q1 (Parent):

SELECT invoice_id, invoice_num, invoice_amount, invoice_date FROM ap_invoices_all

Data Set Q2 (Child):

SELECT invoice_id, line_number, amount FROM ap_invoice_lines_all

Data Sets Q1 and Q2 are linked by invoice_id.

If records exist in Q2 for at least one record in Q1, then the CSV file will be generated with all columns, like this:

INVOICE_ID, INVOICE_NUM, INVOICE_AMOUNT, INVOICE_DATE, INVOICE_ID, LINE_NUMBER, AMOUNT