I'm not quite sure exactly what you are asking as this is an extremely poorly worded question.

Reference databases, while perhaps based on some of the same raw sequence data, are typically curated and manually adjusted for various reasons. I am assuming this is why you have different lengths of sequences in your databases. These are optimizations where data is shortened for speed of query, quality of raw sequence data, and to reduce the size of reference database. I'm not familiar with the ASTRAL database so I don't know how it differs from UNIPROT.