No need to shout, dude 
I can’t imagine a solution in which a centralized entitiy can get as much trust or credibility as a decentralized exchange can get, which transacts utilizing reputed signers, who besides the reputation have collateral to lose, if they don’t behave.
So the reason why BCE is more secure is found in economic reasons as much as in organizational reasons - BCE runs on a blockchain, is decentralized, has no single point of failure that can be attacked and DDoS’ing is is very costly.
Speed is is by design a drawback of decentralized solutions. It takes time to form consensus.
But I wrote that at another place already: unless you want to withdraw funds that were recently traded, I can’t see a reason that prevents recently traded funds from being used for other transactions, like trading them again, before the trade is written in a block.
0-confirmation tx should be possible hwen trading (which speed up BCE by far).
The necessary information stays in the mempool.
It’s only the reputed signers that can mess things up (whi have strong incentives not to do it) and not the users.
If anything goes pear-shaped, some trades are reverted and the users have their original funds back. Only withdrawing funds before a sufficient number of confirmations are reached, must be prevented.
