If they can split an image into multiple tiles at different addresses (as they did in some examples shown in the article), shouldn't they only need to care about the total tile number? Why did they bother to try to fit these postures (more or less) as whole on sheet, or making these sheets at all?
Because they were manually drawing and designing the sheet layouts. If this was an automated process, as shown for the CPS-2 titles, then they could (and did) automate it.
If they can split an image into multiple tiles at different addresses (as they did in some examples shown in the article), shouldn't they only need to care about the total tile number? Why did they bother to try to fit these postures (more or less) as whole on sheet, or making these sheets at all?