Well I think that the key to understanding this is to realize that money supply as other macroeconomic aggregates must be defined over period of time.
If you deposit 100\$ bill at a time $t$ with 10% reserve system money supply can increase up to 1000\$ because first bank keeps 10\$ and lend outs 90\$ next keeps 9\$ and lends out 81\$ etc. and sum of this infinite sequence with 10% reserves is 1000.
You are completely correct to say that if people would start withdrawing money let’s say at time $t+1$ money supply would contract. So in this example at $t-1$ money supply would be 100\$, at time $t$ assuming no borrowing constraints and perfect markets money supply would be 1000\$ (because of the multiplier) and at time $t+1$ when you withdraw the money (so there is run on a banks with this one depositor) then money supply would contract back to its original 100\$.
People keep having this misconception about money supply, that in only increases if there is some permanent irreversible increase in quantity of money but that not correct. For example, consider another macroeconomic aggregate - unemployment, you would probably say unemployment increased if a person lost job even if in a year or so that same person gets their job back. The same applies to all aggregates.