patternsearch - ignoring TolFun

I have a detailed function which I'm trying to optimise. The first step is to use patternsearch from the global optimisation toolbox. Within the options structure I've passed in parameter values TolX = 0.01 and TolFun = 0.01. I've verified that these are indeed the values within the option structure at the point where patternsearch is called. However, this is the information it then outputs:

Iter     f-count          f(x)      MeshSize     Method
    0        1   2.78578e-007             1      
    1        2   2.78578e-007           0.5     Refine Mesh
    2        6   2.78578e-007          0.25     Refine Mesh
    3       10   2.78578e-007         0.125     Refine Mesh
    4       14   2.78578e-007        0.0625     Refine Mesh
    5       18   2.78578e-007       0.03125     Refine Mesh
    6       21   1.56899e-007        0.0625     Successful Poll
    7       25   1.56899e-007       0.03125     Refine Mesh
    8       30   1.56899e-007       0.01563     Refine Mesh
    9       35     5.555e-008       0.03125     Successful Poll
   10       39     5.555e-008       0.01563     Refine Mesh
   11       44     5.555e-008      0.007813     Refine Mesh

Optimization terminated: change in X less than options.TolX.

The values of f(x) (let alone the changes in f(x)) are well below the tolerance level, but the optimisation doesn't terminate until TolX is breached. Does anyone have any ideas as to why this would be the case?

Answer 1

[Grzegorz Knor]

Algorithm stops when change in the objective function in two consecutive iterations and the mesh size are both less than Function tolerance.

So result is correct, in 11 iteration change in the objective function and mesh size are both less than 0.01.