Ambiguous saddle calculations

[smallsaucepan]

Hi there. Seeing the below issue where a particular cell is being treated ambiguously. The data grid is:

  [9, 9, 5, 5, 5],
  [9, 9, 5, 5, 5],
  [5, 5, 9, 5, 5],
  [5, 5, 5, 5, 5],
  [5, 5, 5, 5, 5],

and there is a break set at 7. The two rings returned are show below. The cell marked with a red 'x' seems to be getting treated differently depending on which band is being calculated.

Increasing the break to 7.01 fixes(?) the left hand band:

Decreasing the break to 6.99 though has no effect. Any ideas? Is this the expected behaviour?