Is there any data comparison if the transaction size after upgrade in this case could be bigger than before?
Not exactly data compression but more like data omission. In Taproot we get rid of some extra (useless) bytes:
1- Removed DER encoding used by ECDSA signatures
2- Changed how last byte in signatures known as SIGHASH_TYPE works (it can be removed to represent SIGHASH_ALL)
3- Dropped the first byte of public keys indicating y coordinate oddness/evenness
4- By using Schnorr signatures (ECSDSA) we can now easily have aggregated signatures and keys which means even with multiple participants (such as multi-sig) we still need a single signature + a single public key
The first 3 are affecting all types including the "wallets with only 1 signature" ie. the simple key path spending rules and the last one affects more complex scripts ie. script path spending rules.