I have to agree with both your observations on this. My belief is that it is more to do with winds entered on the descent profile not matching with the actual encountered winds during descent. Hence, the calculations at time can be slightly off at times. Have you guys tried to back track the STAR and add pseudo constraints onto the profile (while ensuring those contracts are complaint with the published constraints) ? I have noticed when this is done a better descent profile is flown by the FMGC. However, I truly believe the DPO update is better than its predecessor.