PL5E raw demosaicing quality

Camera Information Capacity for Machine Vision and Artificial Intelligence systems P. 1 Camera Information Capacity: A Key Performance Indicator for Machine Vision and Artificial Intelligence Systems

Table 2. Capacity losses for sharpening & noise reduction
Image C MTF50 (c/p) MTF50P(c/p)
Baseline 3.69 0.22 0.229
USM R2 A1 3.65 0.345 0.323
USM R1 A2 3.63 0.407 0.397
Gaussian 0.7 2.99 0.162 0.168
Gaussian 1.0 2.25 0.138 0.143
USM R2A1, Gaussian 0.7 3.06 0.241 0.239

Table 3. Comparison of demosaicing algorithms
Demosaicing algorithm C (bits/pixel) MTF50P (C/P)
dcraw bilinear 1.8 0.191
dcraw VNG 2.51 0.219
dcraw PPG 3.14 0.229
dcraw AHD 3.69 0.229
RawTherapee AMaZE 3.95 0.236

Information capacity C is a far more sensitive measure of demosaicing quality than MTF50P.

End excerpt.

This probably is a question for DxO engineering. How does PL5E DeepPRIME compare on the above measure for a supported raw image, in particular for demosaicing quality?