This doesn't save as much (of anything in particular) as you'd hope. The problem is that you are comparing the pathologically bad case of bitcoin to the pathologically good case of the alternative.
Enlighten me. Bitcoin smallest transactions are ~250 bytes. Tx signature takes ~70 bytes. With account ledger smallest transaction would take 24 + 70 = 94 bytes while tx to address not currently in db would take ~22 bytes more and be still 2x smaller than bitcoin. Average tx would probably beat bitcoin even more (no need to combine inputs)
Second, a ledger system is necessarily less flexible and useful. None of Mike's contracts would work in a ledger system without adding trusted third parties. You could regain that functionality by emulating the current scripting system with the scripts attached to disposable accounts rather than disposable transactions, but why bother? Bitcoin puts the complexity where it belongs, in the transaction.
Well bitcoin is more flexible only in theory. Most functionality of scripts is turned off until developers decide such scripts are nessesary and won't bloat blockchain, network etc. It's effectively not more flexible than account ledger.
Bitcoin scripts are constrained while with ledger you can make more powerful concepts because you have access to turing complete programming language and current network state. You can make complex constructs for example to make secure atomic laundry (
https://bitcointalk.org/index.php?topic=195275.msg2289285#msg2289285) and much more.
This is perverse logic. You are again comparing the worst case in one system to the best case in another.
These were description of capabilities. No logic behind it. Programming language is more capable than scripts simple as that. Scripts are more flexible in theory but in reality they must be artificially limited to prevent abuse.
EDIT: Nice recent example of how scripts degenerated in 'convince developers your script use-case is important'
https://bitcointalk.org/index.php?topic=220273.msg2320603#msg2320603).