In my book I explain that you have to preprocess the data and create one giant record for each check. The simpliest way is to create a cursor with all the fields for the check and memo fields for each column in the stub. For example, have one memo field for the date, one for the description, one for the amount and so on. When preprocessing the data, fill the memo field with the first detail line's data, followed by a CHR(13), followed by the next detail line's data, followed by a CHR(13) and so on.

When designing the check you can make the detail band be as big as the entire page. Call a UDF that splits out the memo fields to arrays (=ALINES(aInvAmt, tmpCursor.InvAmt). Stack a bunch of fields in the detail band for stub 1 and stub 2, each with the expression of aInvAmt[1], aInvAmt[2], and so on.