I do not think you should use really big corporations as example.
First of all, they attract enough talent to use any methods of selection they see fit. They do not care about false negatives, since they operate on virtually unlimited pool of resources.
Secondly, the logic against whiteboard interviews is sound. I have never had to implement complex algorithm at work. Never. Which means that last time I actually wrote some was at college, about 7 years ago. Obviously, to perform well, I need to study to remind myself how to use red-black trees for example. Or how to implement priority queue (hopefully not using Fibonacci Heap, ugh). Any effort that goes into it is basically wasted, since chances I will have to actually do this at job are close to zero. So I'm basically reminding myself how to do something I will never do past interview. Not very effective.
Third and last, good software engineering is, as I believe, not strongly correlated with algorithms knowledge. People great at hacking quite often get in trouble when they have to develop readable, predictable system. This is different set of skills.
Secondly, the logic against whiteboard interviews is sound. I have never had to implement complex algorithm at work. Never. Which means that last time I actually wrote some was at college, about 7 years ago. Obviously, to perform well, I need to study to remind myself how to use red-black trees for example. Or how to implement priority queue (hopefully not using Fibonacci Heap, ugh). Any effort that goes into it is basically wasted, since chances I will have to actually do this at job are close to zero. So I'm basically reminding myself how to do something I will never do past interview. Not very effective.
Third and last, good software engineering is, as I believe, not strongly correlated with algorithms knowledge. People great at hacking quite often get in trouble when they have to develop readable, predictable system. This is different set of skills.