Maybe, but attacks on anonymity can be quite subtle, with various combinatorial, timing, and sybil type attacks, so I wouldn't be so confident. If you look at unlinkability, untraceability, and amount hiding as three prongs of resistance to blockchain analysis, then he's entirely missing one prong, which makes his argument quite weak. Blockstream has stated likewise about CT not hiding what they call transaction metadata, only content. Stealth is a nice convenience feature, but largely similar to just having good address reuse practices in Bitcoin (which can also be achieved via payment protocols and HD address chains). To do this well, you really need another piece, at least some sort of good coinjoin/coinshuffle type solution, and that is really hard to do well (potentially impossible) given sybil and timing attacks. At least Dash tries, but Larimer dismisses the problem too easily.

So I'd characterize Larimer's argument as largely wishful thinking and/or hype (i.e. this is what I have therefore this is what is needed, the marketers variation of the arguing from the conclusion fallacy).

But that's an entirely different argument from whether strong privacy/anonymity/fungibility (it is very hard to separate any of these from the others) is more important than scalability (or vice versa). I suppose you could also make that argument that without all of these things you don't really have a very strong solution overall and again are engaging in wishful thinking (which was in many ways the premise of TPTB's original Ion "Bitcoin killer" concept, before he neutered it).