[kqr]

If the main reason your effort estimations are off is that you don't allow for unknown unknowns, then that's easily fixable.

After all, although we don't know which the unknown unknowns are, the possibility of them is known. And they do, in my experience, tend to increase the required effort by, say, 1--30 × depending on task complexity and familiarity.

So even in the most complex and unfamiliar of tasks, you can adjust the upper end of your estimate by 30× and there you go! Unknown unknowns accounted for in your effort estimation.

(Simpler or more familiar tasks require smaller adjustments to their upper end. Knowing how much adjustment is appropriate is a matter of deliberate practise.)