The article uses some incorrect terms. A 51% attack implies the attacker has a majority of the hashrate. The article deals with situations where the attacker has a minority of the hashrate (and a very dubious example of exactly half the hashrate). In that article every reference to "51% attack" should be replaced with "double spend". You confuse the issue further by using the term timewarp attack which is one possible exploit for an attacker who has a majority of the hashrate. Network halting, timewarp exploit, and double spending are all exploits that are possible when a malicious entity abuses a situation where it has a majority of the hashrate ("51% attack").
The math in the article while not wrong is misleading. The article implies that for an attacker with <=50% of the hashrate to build a longer chain after one week would be very difficult. Well no kidding. For litecoin one block week is 24*24*7 = 4,032 blocks.
Yes with 4,032 confirmations one can be reasonably sure it is nearly impossible for an attacker with a minority of the hashrate to successful double spend that. The key word is minority. The comparison to Bitcoin is equally dubious because it compares a scenario where a Bitcoin tx has less confirmations (1,008 vs 4,032). Yes once again having 1,008 confirmations is less secure than having 4,032 confirmations when dealing with an attacker who has a minority of the hashrate. The key word again is minority.
The bitcoin pdf has proper although simplistic math for the probability of double spending with a minority of the hashrate
http://bitcoin.org/bitcoin.pdf. Meni wrote a more comprehensive paper that deals specifically with the economics of double spends
https://bitcoil.co.il/Doublespend.pdf. If the attacker can maintain a majority of he hashrate then the success rate of a double spend is 100%. No amount of confirmations can provide a confidence that the transaction can't be reversed. The longer the attacker works on his chain the higher the probability that it will be longer than the "legit" chain. For a short period of time the legit chain may remain ahead due to "good luck" but eventually the attacker will pull ahead.
Or stated another way, if the attacker has a majority of the hashrate then it is a mathematical certainty the attacker will eventually produce the longest chain.