I'd analyse the above graphic (#218) this way:
(throughout bearing in mind reaction time / inertial delays)
Prior to A:
Establishing good glidepath but then offsetting low due to tracking radalt.
From A->B:
Rate of Descent stabilised again: similar to intended glidepath, simply offset low. However radalt tracking falling ground so implies approx zero RoD, 100m alt.
B->C
Response time to second '100m' call.
From C to CFIT:
i) Mistaken attempt to remedy apparent zero RoD. Thus trending to new descent rate: 'the ground incline at around 20s' plus 'intended RoD' (leading to maximum RoD: at about 13 seconds)
ii) Excessive RoD becomes immediately apparent despite 'trust-the-radalt' mindset as terrain is now rising, but combination of:
Glideslope RoD
plus
RoD compensating for previous terrain downslope
plus
effective RoD from terrain upslope
adds up to:
Game Over.