You're definitely right that merely detecting tumors earlier doesn't always help, because many of them are relatively slow-growing and fairly harmless. But this actually makes a very specific kind of early detection research even more important: research to help us predict which early-stage tumors will progress to becoming deadly, and which ones won't. Currently we're pretty bad at making this prediction for many kinds of cancer, but if we could detect them all early AND distinguish the really bad ones from the mellower ones, we'd know exactly who to treat aggressively.