The difference between "z= f(x, y)" and "F(x, y, z)= 0" is that you are taking F(x, y, z)= f(x, y)- z. Taking the gradient [itex]\nabla F= <f_x, f_y, -1>[[/itex].
It is correct that [itex]F(1, -1, 4\sqrt{2})= \frac{1}{4}+ \frac{1}{4}+ \frac{32}{64}= 1[/itex] but that does NOT say "z= 1". You are told that at this point [itex]z= 4\sqrt{2}[/itex]. The fact that [itex]F(1, -1, 4\sqrt{2})= 1[/itex] just verifies that the given point is on the given surface.
The tangent plane is [itex]z- z_0= f_x(x_0, y_0)(x- x_0)+ f_y(x_0, y_0)(y- y_0)[/itex] only applies if you are given z= f(x, y) which is NOT the case here. You are given, rather, [itex]F(x, y, z)= x^2/4+ y^2/4+ z^2/64= 1[/itex]. In order to convert that to the "z= f(x,y)" form you need to solve for z: [itex]z^2/64= 1- x^2/4- y^2/4[/itex], [itex]z/8= \sqrt{1- x^2/4- y^2/4}[/itex], [itex]z= 8\sqrt{1- x^2/4- y^2/4}[/itex].
Much simpler is to use the fact that if F(x, y, z)= 1 (or any constant) then [itex]\nabla F[/itex] is perpendicular to the surface at any point. Since for any point (x, y, z) on the tangent plane to the surface at [itex](x_0, y_0, z_0)[/itex], the vector [itex](x- x_0)\vec{i}+ (y- y_0)\vec{j}+ (z- z_0)\vec{k}[/itex] is in the tangent plane, we must have [itex]\nabla F\cdot [(x- x_0)\vec{i}+ (y- y_0)\vec{j}+ (z- z_0)\vec{k}]= 0[/itex] which is the same as [itex]F_x(x_0, y_0, z_0)(x- x_0)+ F_y(x_0, y_0, z_0)(y- y_0)+ F_z(x_0, y_0, z_0)(z- z_0)= 0[/itex] as the equation of the tangent plane.