Hi Dmitry,
I'm assuming by rectified you mean the precision of the resampling/remapping process? I don't fully understand the math involved in the algorithm, but I can tell you what I do know.
The default resampling for Polar2Grid uses an EWA algorithm which stands for "Elliptical Weighted Averaging" and consists of two steps. The code for this resampling was originally based on the "ms2gt" software created by the NSIDC (
https://nsidc.org/data/modis/ms2gt/index.html). It was rewritten to work better with python, but the overall algorithm is still the same. The first step is called "ll2cr" which stands for "lat/lon to col/row". In Polar2Grid this uses PROJ.4 (via the python library pyproj) to navigate each longitude/latitude pixel to a cell on the output grid. The second step called "fornav" for "forward navigation" is where the actual EWA algorithm is performed. This algorithm operates on one scan line at a time (the whole swath for AVHRR) and calculates some coefficients and weights based on the location of each pixel in the scan. It then looks through each input swath pixel, applies the pixel to each output grid cell around it (in an ellipse), applying a weight based on how far the grid cell is from the swath pixel. Each grid cell will have the sum of the input pixels that affect it and the sum of the weights applied to those values. It can then perform the average "output = input_sum / weight_sum".
So the parameters involved are the longitude and latitude of each input pixel, the actual swath data, and the number of input data rows making up one scan line (whole swath for AVHRR). TLE files are not used. The EWA algorithm does allow you to control some of the coefficients (effect radius of each input pixel, etc), but P2G's defaults were chosen to produce the best results so you probably don't need to change these.
That might not be the answer you were looking for, but I hope it helps.
Dave