You're getting NaN values from GRIDDATA because some query points are outside the convex hull of the sample data points. GRIDDATA only supports interpolation within the convex hull. The following piece of code shows the sample data in blue and the query points in red.
[X,Y] = meshgrid(x,y);
X = X(:);
Y = Y(:);
dt = delaunayTriangulation(X,Y);
plot(a,b,'.r')
axis equal
hold on
triplot(dt);
hold off
You can use extrapolation to approximate the values outside the convex hull. scatteredInterpolant is similar to GRIDDATA, but in addition it supports extrapolation. This sample data lies on a regular grid, so the grid-based interpolation algorithms are more efficient to use here. You can use griddedInterpolant to solve this problem and this also supports extrapolation outside the domain of the grid. For example, use the following code instead:
[xq,yq] = ndgrid(x,y);
F = griddedInterpolant(xq,yq,e');
[a,b] = ndgrid(m,n);
c = F(a,b);
