Nobody of those says that SICP isn't a really great book.
Wadler claims that the book and a 'better' language would be an improvement.
Felleisen et.al. claim that the book does not prepare the students for software engineers in industry, is very difficult and thus is not the best book for a typical introductory course.
Even though I understand both critiques, I see their point of view and find both views valid, SICP is still * out of five stars.
The first one is by Philip Wadler, who's quite famous in Haskell circles. The second one is by Matthias Felleisen, is quite active in using Scheme to teach computer science.