Yes this is very true; however most of what Bitcoin replaces in the fiat world is not cash (M0) but funds in bank accounts, credit on credit cards, funds on deposit with Paypal or a similar service, money market funds etc. M1, M2 or higher.
With cash, there aren't many options to accumulate wealth by securing cash at a residence for example.
At some point cash can no longer be kept secure without additional costs and risk exposure, and the currency may be getting devalued to boot.
The options then include investing it as capital into a business, lending it to another party who will invest it in a business or other party that pays interest, or even converting the cash to another asset (e.g., precious metals) and storing wealth there.
With Bitcoin though, your currency isn't at risk of being devalued and it can be kept secure at minimal cost (when done properly). So while many people have the mindset that they need to put their bitcoins to work earning interest, an even larger number of others will be content earning no interest on their stash of coins, just sitting on them until there is a good use for them.
Thus bitcoin currency overall can have a lower velocity (how many times in a period the currency changes hands) than other currencies.
Offsetting that tendency though is how when bitcoins are used for consumption, commerce, investment, lending, etc., Bitcoin's properties (nearly instant for payment notification and settlement, no chargebacks, useful globally, etc.) all allow the currency's velocity to skyrocket versus any other currencies that might involve the banking system or are restrained due to the laws of physics (delivering cash, in-person, to a supplier).
While fractional reserve may eventually happen in the Bitcoin economy.
It's already here. I suggest to set a date for "Anti Fractional Reserve Day". During this day everyone should withdraw his/her coins from exchanges, webwallets and similar services to a personal wallet. And deposit the coins back in 24 hours. This "trick" will definitely help us to keep Bitcoin economy "healthy".
That's why the category of deposits called term deposits (also referred to as time deposits in the U.S., such as certificate of deposit / CD) and bonds will likely need to be more a widely used tool in the Bitcoin ecosystem so that lenders are less vulnerable to bank runs versus if they had only sight deposits (on-demand withdrawals).
It would be trivially easy for an exchange or webwallet to be able to prove the amount of BTC they hold but unfortunately doing so reveals the deposits and withdrawal payments, which in turn violates customer privacy. The webwallet problem essentially can be solved with a hybrid EWallet like Blockchain.info/wallet though that requires the account owner to configure the account properly. For many, their chance of losing their funds by not properly configuring their hybrid wallet correctly (giving it an alias, getting backups to dropbox or e-mail, etc.) might exceed the risk of a trusted hosted (shared) EWallet absconding with the funds or getting hacked.
As far as exchanges, what SMPake is doing might help lessen the need to keep bitcoins at an exchange. Many traders today store coins at an exchange as that is the only way to have a sell order remain active. Exchanges needs this so that they can do instant settlement on each trade. But if there is the ability to transfer funds in one block due to an intermediary like SMPake, then a trader needs to store fewer coins at any exchange but still is able to put up sell orders in a timely fashion.
That harms perceived liquidity though so maybe exchanges will be pressured to consider another approach.
Using M of N multisig, e.g., 1 of 2, the exchange can spend my funds even though they don't store them. The problem is there isn't instant settlement with each trade -- the exchange is exposing themselves to the risk that the seller will issue a double spend (e.g., Finney attack to invalidate the trade). But the exchange can counter this by simply restricting access to the fiat funds when that happens until the blockchain transaction for the trade has fully confirmed.
There are so many solutions available. The problem is with so little competition, exchanges can follow the same current obviously flawed model that they currently use and make a little profit while passing on the cost of security to its customer base (in the form of direct losses resulting from theft). Until the customers demand more, nothing will change.