I can give you the answer to that question.
The removal of any signal requires the production of a proper signalling design.
To do this requires that the existing signalling equipment on site and in th relay room is surveyed, a process known as "Correlation". The Correlation is against existing signalling drawings compared to the actual situation in the lineside location cases, relay rooms, cable routes, etc, etc.
The various re-organisations within BR and then subsequently Railtrack / Network Rail have resulted in the loss of considerable amounts of Infrastructure records, indeed in some cases, we do not have any accurate picture of what is under the track when we go to do Renewals in station areas, other than from local knowledge, which is now fast disappearing with NRS policy of removing old BR people.
A similar situation applies to signalling. This can be compounded by maintenance alterations which are often only recorded on site and never transferred into the principal records.
On site, many location cases were installed many years ago and can suffer from a degradation to the rubber insulation on the cables, this is known as either "Wet Wire Degradation" or "Dry Wire Degradation" depending on which type is involved.
Similar problems occur in signalling relay rooms, and Railtrack was certainly going forward with the renewal of such wiring.
This can mean that it is not sensible to disturb any cables in location cases because the risk is that the insulation is damaged which would then require a complete wire change, which in turn could lead to damage to other cables and so on.
Where no records exist, then a new record would need to be produced from scratch and would require considerable time and drawing office effort just to replicate the former records.
The next stage is the revised design.
Now obviously the ground position light signals will be wired into various signalling circuits, and their removal would require alterations to these signalling controls together with alterations to the Control Tables, which in turn brings with it a requirement for full Signalling Principles Testing.
Signalling design and testing resources are scarce, and as a result they are expensive.
You can see therefore that what on the face of it appears to be a very simple task (just cut a few wires and remove a signal) is actually a much more complex engineering task. Far better therefore to leave the signalling controls in situ rather than recover.
Now on the LNE Zone, we developed a process known as mimimal signal recoveries, which set out what was left and what was recovered. In most cases we recover junction indicators and calling on signals mounted on main posts, but leave ground position light signals. The wiring in the location cabinet for these signals is "turned round" , i.e. a strap is placed across the controls to effectively short circuit the bulb detection system, the outcome of which is to "fool" the relay room into thinking that the signal is permanently alight and displaying a red aspect. The power is taken from the signal, and in some cases the signal is recovered, but not always. Do you want to waste money on a recovery that may not be necessary is the question.
In some cases the wiring may not be in a good condition in the location case in which case the signal will be left alone, and this can result in the sorts of instances you have identified.
In a world where money was no problem, then we would correlate everything, bring the records up to date, do the redesign, recover and test the recoveries, implement the new controls, recover the redundant signalling equipment and controls and update to final the main records, with copies redistributed to each Maintenance area, and fresh copies of the drawings into the location cases.
It really IS down to how best to spend money, and I will always counsel for minimal recoveries as the most cost effective way forward, given all that I have explained above.
Finally in the case of new installations it is much cheaper to install equipment in advance rather than return and do a redesign. It may look foolish, but it is cheaper in the long term.