A lot of science work consist of what my scientist mother calls "bread work" - validation of existing techniques, building a general corpus of knowledge, maintaining databases with useful information and so on. These is weak-link work. A bad quality database will undermine any research which is built on it and not aware of its limitations.
A good recent example may be the stem cell research field, which was (and perhaps still is?) bogged down by being built on bad fundamentals.
But yes, more broadly speaking I do agree that science funding should be built to encourage more risk of failure, more elitism, and more consolidation of large project to be able to "punch through".