Can I get good results with the scanpy ingest function when the referenced dataset (2638 × 1838) is relatively small and my own dataset is relatively large(18929 × 2664)? When I annotated my data with ingest and after, the dataset becomes similar to the reference dataset.