Exactly, and imagine the credibility of a 'fork team' who on one hand is trying to sell their fork as the 'new Bitcoin' that will 'replace it' (flippening), while at the same time being based (thus reliant) on the real BTC blockchain through merged mining. Nobody would take them seriously.
There is a binary decision which chain to mine, that doesn't exist in PoS.
It is partially true. There is a binary decision which chain to mine, because you have to choose your previous block hash of your block header that you are trying to mine. But if Merged Mining is active, then you can reuse the same header on both chains. Then yes, it is clearly visible, which chain you picked (because of that previous block hash), but the reusability of that header makes things different, when it comes to the profitability.
Imagine a scenario, where you picked BTC as the chain to mine, but mined a 100 times easier block header. Then, if Merged Mining is implemented in a NameCoin-like way, as a separate difficulty, you will get around 0.0625 BTC plus fees, and for example 50 ALT. But, if the altcoin is constructed to collect all headers, and adjust the coinbase, instead of adjusting the difficulty, then the whole profitability is different, because then you will only get around 0.0625 ALT, and the rest will be pushed to the future.
So, while creating separate difficulties is not serious, collecting all headers, valid or not, and adjusting amounts, is a different matter. Because then, even if there is a stronger chain, and if those miners could potentially switch their machines to attack the chain, the network is aware of the potential attack, and things are stated explicitly, because the total hashrate is always tracked and reflected in coin amounts.
So, to sum up, if someone is seriously thinking about any ALT that could "replace" BTC, then such chain should be at least work-per-amount-compatible, so the total effort to mine 1 BTC should be comparable with the total effort to mine 1 ALT.