If you want to end up storing the result digitally then there is no point in this step (and the steps leading to this) because the main purpose of encoding the result as a set of words is to make it easy for the user to write down something they see on their screens with minimal risk of mistakes.

This brings up another problem with your algorithm. Having a large word list means in that list there are words that ought to look similar (act, apt, aft,...) which means the risk of typos when user writes it down is high and the reason for using a word list is not there anymore.

Another problem is the fact that you limited this algorithm to 128-bit entropy/seed whereas BIP39 entropies can be from 128 to 256 bits.

And biggest problem of all is that this is not exactly a cryptographically strong encryption algorithm. A better method would have been to use an actual encryption algorithm like AES to actually encrypt the entropy/seed and then encode that result however you liked and then stored that instead.