I don't know what the spreadsheet is calculating, but here is what the "Surface GF" is that recent versions of Subsurface show in the info box (the code is here:
Subsurface-divelog/subsurface ). Every model I am aware of is calculating tissue inert gas loadings. The natural units for those are partial pressures, something like 1.53bar for the partial pressure of N2, say. The problem is that this number alone is not very telling (is it high? or low?). So, for the user to judge it is worthwhile to translate that to other units or to set it in some relation to another pressure.
To me, it appears to be most understandable to translate this to a ceiling depth, i.e. to the minimal depth you could ascent to with this tissue loading. And in Subsurface, we show this as the green ceiling areas and numerically in the info box. But if you insist on showing it as a pressure rather than a depth, a good reference pressure to compare to is the m-value, the (according to a model) maximal allowed tissue pressure at a given depth. It has become conventional not to compare the absolute pressures but rather the excess over the ambient pressure, i.e. compute the ratio
(p_tissue - p_ambien) / (m-value - p_ambient)
If you use plain vanilla (GF 100/100) Bühlmann to compute the m-value and express the ratio as a percentage, this is "the gradient factor". Note that since this involves the ambient pressure which is increasing with depth and also the m-value in itself is depth dependent, the numerical value of the gradient factor depends on depth. But we were after an absolute measure of the amount of gas in the tissue and since the amount is not depth dependant, we should better use a value that does not change when you change depth (without changing the amount of gas in the tissue). So the idea off Shearwater was to use the "surface GF", i.e. the gradient factor that you would have if you used depth=0 and ambient pressure = surface pressure in the above calculation.
So naturally, if you are at the surface, the gradient factor and the surface gradient factor agree, but at depth they don't. My guess is that this is the difference (the spreadsheet is computing momentary GF). To repeat: the (momentary) gradient factor tells you how much of the allowed (according to 100/100 Bühlmann) excess pressure over ambient pressure you are currently using, it is a measure of the momentary deco stress if you like. On the other hand, surface GF is a measure for the absolute amount of gas in the tissue, independent of depth.
Subsurface does not numerically display momentary GF (as I think the ceiling depth gives a better representation of the tissue loadings), but it is shown graphically in the tissue loadings bar graph.
So, independent of maxdepth of your dive, at the end of your NDL time, the surface GF will approach GFhigh. This is by definition (of "NDL"), as the NDL is determined by the fact that you don't occur any mandatory stops and you would need to do a stop if without the stop you had a GF at the surface exceeding GFhigh.