@Alkerion Have you tried the M1005 [In] gcode format for skew correction? I find it gives better results than the M1005 [Xn] [Yn] method. The M1005 [Xn] [Yn] format iteratively applies a new compensation factor to the previous one and can make the results jump around a lot unless your measurements are very accurate. M1005 [In] overwrites any previous correction.
As suggested by @lindnjoe uniformly scaling the test print larger will give you more accurate results.
See this thread: Skew compensation in the Bambu X1C Firmware for details of how to use gcode M1005 with the [In] parameter.