There is differential and differential. My Garmin 296 uses EGNOS to provide differential, Europe -wide. In the States the 296 will use WAAS, a similar but completely separate differential system, to achieve the same thing.
Offshore we used multiple reference DGPS using differential stations around the North Sea to provide horizontal accuracies of slighly better than +/1 to 2 metres. The bees knees uses something called Precision Point Positioning (
www.veripos.com/ultra) which does NOT use differential but software modelling to eliminate inaccuracies and achieves +/- 0.2 metres or better horizontally. Slightly worse vertically but certainly sub-metric. More usefully the precision is achieved worldwide rather than in a constrained area. DGPS signals decrease in accuracy the further from the control station, less than 1000 miles would be normal.
If you have been using a GPS to provide OSGB coordinates then I would expect that the software would convert the spheroid from WGS84 to that used in the UK - Airy (1936). That is the height would be converted to OSGB values which are based on conventional spirit levelling. Although the plaque was installed in the 1970s the OS achieved astounding accuracies through out the UK using techniques that would now be considered too expensive and labour intensive.
However it is unlikely that the AA went through the same process to survey in the plaque..they may have used an OS benchmark nearby but then again they could have looked at a map and taken the elevation off that.