My understanding is that the idea behind CD NPA is to try and make the execution of these as similar as possible to an ILS.
The flaw is that it pretends a little too much that the non-precision approach is an ILS , i.e. a precision approach.
If the onboard equipment (e.g. BaroVNAV) allows you to fly to a point in space with high vertical and horizontal precision, then you might as well treat it like a precision approach. And because you dip (because you're treating it like a DH) below the MDH/OCH at that precise point, where there are no obstacles, there's little risk of hitting something even if you're in still cloud.
But if you're using raw data, selecting a rate of descent with mile-spaced DME crosschecks or worse, there's a significant variability in the distance from the runway at which you hit the MDH/OCH, and your dip of 50 ft eats directly into the margin above obstacles on the approach. I would love to see the risk assessment that demonstrates that dipping 50 ft below the OCH in these circumstances is within tolerable risk.
The progression towards CDFA for high-inertia, multicrew commercial air transport with appropriate avionics is clearly a step forward in safety. But the one-size-fits-all attitude of EU-OPS is not.