That seems correct to me, but I haven't actually tested to see if it is statistically accurate.
The other way it might work is that the improvements are stacked. So with the fortress you end up moving from an 8 to a 12 (50% improvement), but then the 25% actually stacks on the *new* score, so you move to 15 (25% increase in 12). This is opposed to a 75% increase in 8, which would lead to 14. The stack method is obviously better for the person with the improvements. However, I think the simpler method of adding the percentages is probably used.