It only means you need more blocks for the same security
Completely false. The probability that a fork will succeed against an attacker with X percent hashpower after k confirmations is (X / (1 - X)) ^ k. There is no term for block time in there. You could argue that with a shorter blocktime an attack is cheaper so an attacker has more chances, eg. with 6s block times an attacker has 100x more chances, but that effect is linear, whereas the effect of adding more confirmations is exponential - assuming 30% hashpower, waiting for 12 confirmations instead of 6 reduces the probability of a fork succeeding by a factor of 161, so you get the same security in 72 seconds that Bitcoin gives you in 3600.
Agreed. If one reads the original bitcoin paper, they'll see that block time is not a factor in the main calculations about the security of confirmations.
Shorter block times do have some negatives (higher chance of blockchain forks which means that miners waste more time not helping secure the network, and miners with faster connections get more of an advantage over miners with slower connections because they waste less time on outdated blocks), but the security of confirmations certainly does not scale linearly with the block target time.
Anyway, it's overkill to wait for even a single confirmation for most small transactions. You can wait a few seconds to see if a double-spend starts trying to spread across the network. And in many situations where people pay with BTC, like subscribing to a service, or ordering something to be shipped to them, there's no need to make the customer wait around staring at a screen telling them to wait for confirmation. If a double-spend happens and the transaction fails, then the seller can cut off the user's subscription or stop preparing to ship the item, and then contact the customer.