Alignment with MTCNN does not produce insightface package result

Hi,

I have one image named sample.jpg and it contains only one face. I detected and aligned the face using mtcnn and parameters you have mentioned [here](https://github.com/deepinsight/insightface/issues/1286). Then I used `insightface.model_zoo.ArcFaceONNX` to get the aligned face's embedding. Then I used `insightface.app.face_analysis.FaceAnalysis` (large model) to detect and get embedding of the sample.jpg image. But comparing two vectors results **0.9034** similarity and **0.1410** distance which should be 1 and 0.

Any idea?

Update:
I changed the alignment parameters to those in [face_align.py](https://github.com/deepinsight/insightface/blob/master/python-package/insightface/utils/face_align.py) in insightface package and got **0.9690** similarity and **0.0793** distance but they are not 1 and 0 yet. Is it normal? because detection models are different (MTCNN and SCRFD)?