You will need experiment with your WEEDING factors, and the other surface creation factors, (smoothing, face flipping, etc.) And you may never get them to match exactly, (with any product). As you are asking an algorithm to return the same resultant from different data. Those contours are the interpolation of the original data, (points, break lines), and the data extracted from them is rarely ever going to match exactly. You can get close; just keep the above in mind.
You might post your data here and we can poke at it to see just how close we can get.