It's very easy to accidentally get misleading benchmarking results in 100 different ways, I wouldn't assume they did no benchmarking when they did the duplication.