findEssentialMat threshold issue

I find a peculiar issue, that to use the findEssentialMat, one has to normalize the points based on the discussion here, by multiplying the inverse camera matrix. However, in reality, we could undistort the points first and then still use focal = 1 and pp = (0, 0) as if the points are already normalized, even though they are not, as reported here and here.

The I come across the code for monocular VO in paper named Autonomous Visual Mapping and Exploration With a Micro Aerial Vehicle here.

It includes both solve5PointRansac and solveP3PRansac. The solve5PointRansac uses findEssentialMat and recoverPose. The solveP3PRansac seems to use the RANSAC implemented by the authors.

Note that in motion_estimation/npoint/fivepoint/src/main.cpp, the focal passed to findEssentialMat is 300, while the threshold is 1.0 - 1e-10. In the findEssentialMat, the threshold is changed according to the focal as threshold /= focal; therefore it seems that we could not normalize points first and then use focal = 1, as this would leave the threshold unchanged.