Fisheye lens distortion correction and points extracted from checkerboard

Hello, I have two questions. 1. Fisheye lens distortion correction works great with the images provided by Matlab. However, with my images, most of them are rejected at the feature detection stage. To find rejected images, change one line in example code: [imagePoints,boardSize, imagesUsed] = detectCheckerboardPoints(images.Files); If I leave only accepted images and run the calibration, the correction coefficients are such that the corrected image does not look pincushion-like (as in example result) but rather barrel-like. This is apparently wrong, as fisheye distortion is barrel distortion. What I might be doing wrong? 2. The calibration app for normal (not fisheye) cameras works fine and gives results close to that of Caltech's toolbox. However, I needed the locations of the detected features, so I saved them in a file. And found that the locations are wrong - they are all shifted by approximately 2 pixels. If this behavior intentional, or is it a bug?

Thank you, Yuri