Not really. The design of Bitcoin assumes that an attacker doesn't control more than 50% of the total mining power. If that's true, the level of protection provided by N confirmations decreases exponentially as N increases and is essentially independent of the confirmation time. This is all in Satoshi's original whitepaper.

If an attacker does control 50-plus percent of the total mining power, you're basically screwed. Increasing the total confirmation time doesn't increase the cost of a 51% attack by enough to be worth attempting.

Well I think the argument is that the reliability is increased by using scrypt instead of SHA256 which makes it harder for someone to control more than 50% of mining and that the result is about even.

In practice it doesn't seem to matter. For example btc-e.com who accepts both bitcoin and litecoin for trading makes you wait 3 confirmations for bitcoin and 6 confirmations for litecoin. So you are correct that 1 litecoin transaction is not as reliable as 1 bitcoin transaction but they are not making you wait 12 litecoin transactions.

Given that btc-e has a lot to lose on double spend attacks I think it's a good measure of how these two currencies are playing out in the market.

The propagation times are very important here. You can get away with faster if and only if the network communications latency is low enough.

If blocks are too fast you start getting very large reorganizations.

Today litecoin has almost no transactions, so its lower interblock time is not a problem... Yet.

How is that? My understanding is all it does is slightly reducing the time it takes to successfully make a 51% attack, and slightly increase the risk of orphaned blocks.

The reduction in time to perform a 51% attack is countered with just a few more conformations.