Thanks for your reply!
By control the delay, I meant I want to, for example, delay one of the output clock signals by 0.1 ns, precisely.
I am under the impression that inserting an inverter could definitely create some delay. however, such delay somewhat depends on the process variation, unmodelled parasitics, etc. Both of them will make the simulation very different from the measurement.
And I am also curious about your comment: "Current starved inverters is a common delay technique. " Is this statement valid for clock design, which requires a relative precision?