Rolling Through put Yield question

Hi,
Apologies if my question is in the wrong place, I'm quite new to this and learning on the job!

I am currently introducing TPY measurments at various stages of manufacturing at work (with rolling TPY's in 2 points).

Most of the TPY points are standard (total passed without issue / total number input). But at one point I can actually measure the total number of defects, therefore can get the Defects per unit.

I have read that TPY can be calculated using an exponential equation with DPU, so tried this. But the results differ greatly from the standard TPY! Would it be considered more accurate to use a DPU based value? Or should I be consistent and use the same method for all TPY's to get the rolling TPY value consistent?