Additionally, there are a wide variety of traditional diets that cover very different foods. Since everything in the body interacts in complicated ways, we cannot even generalize specific foods as being "non-risky" to eat, since the empirical evidence we have only applies to the interactions of each food with the rest of the diet. So it's not clear what would actually constitute a definitive "traditional" diet; the best we could do would be to try and mimic a specific traditional diet as closely as possible, which still doesn't take into account the interactions caused by non-diet aspects of health like amount of exercise.
There is value in that kind of stability, but by incorporating mainstream nutrition research into your diet you can trade increased risk for what is likely to be a better average result. I say likely to be better than average because, as incomplete as nutrition data is, some data is still better than no data. It doesn't make sense to ignore what we know in the moment just because it might be wrong later. As long as you research carefully and stick to the most well studied aspects of nutrition, risk is minimal.
It's also worth pointing out that the normal person's diet today is already a large departure from traditional diets. So even if we assume a "traditional" diet is the goal, it does not follow that that Soylent would be better or worse than the normal person's diet today. It's likely that the human body is adaptable enough to handle whatever you eat.