Excellent points! I admit that I open the article with a rather stridently anti-survey position, but I’ve found exactly the same thing you mention and I think you’re right that you need to use both quant and qual methods judiciously.
Qualitative methods like conversations and observation are generally better for uncovering unknown unknowns – they let you dig deeper, explore needs and problems you never thought of, and find out which of your assumptions you need to challenge.
Surveys are useful in two big ways: to begin to learn how universal or crucial the needs and problems you’ve uncovered are, and to provide evidence that is more powerful at persuading analytically-minded stakeholders.
Throwing away a whole genus of research techniques is a bit like trying to play a piano concerto with one hand tied behind your back. When most people seem to leap to using only (poorly constructed) surveys and tying their qualitative hand behind their back, I feel OK about bashing the survey hand pretty hard :)