There seems to be a scaling variable missing for ‘t’ in the ‘x(t)’ calculation, and your ‘t’ vector does not have fine enough resolution to produce the necessary details in ‘x(t)’.
t = linspace(0, 1, 1E+5);
and you can avoid the loop entirely with:
[T,I] = ndgrid(t, (1:30));
x = cos(2*pi*g.^I.*T)./g.^((2-F)*I);
xs = sum(x,2);
I will let you sort the reason that the plot looks appropriate for (0,0.2) although not on (0,1), the reason I mentioned the ‘scaling variable’.