What I can't get my head around though is the increased Baro Alt minima for a regular CAT I approach, for which the decision point is not determined by reference to the Rad Alt...
That typically means there is an obstacle somewhere along the ILS flight path, at an altitude below the LOC MDA but penetrates the protection surface of the 3.00 degree ILS.
Remember that while the LOC MDA is a "hard deck", the ILS has to provide protection even if the aircraft is flown below the glide slope (up to full deflection), and below the DA when a missed approach is initiated. So near the runway the ILS protection is actually below the MDA in this case.
Additionally, the ILS design has to take into account that a real-life glideslope doesn't precisely signal the ideal flight path but may have some errors up and down due to reflections, etc.
When there are obstacles penetrating the ILS surface, the two basic choices are: 1) raise the DA(H); and/or 2) raise the glideslope angle beyond 3.00 degrees.