Normally, Insertion Sort searches backward through the sorted portion to find the correct position.
Using Binary Search improves this search.
For the $i$th element, the insertion position can be found in: $\Theta(\log i)$ time.
It may therefore appear that the total complexity becomes: $\Theta(n\log n)$
But after finding the correct location, elements must still be shifted to make space.
In the worst case, inserting the $i$th element may require shifting: $i-1$ elements.
$\therefore T(n)=1+2+3+\cdots+(n-1)$
$\Rightarrow T(n)=\frac{n(n-1)}{2}$
Hence,
$T(n)=\Theta(n^2)$
Binary Search reduces the number of comparisons needed to find the insertion position, but it does not eliminate the linear number of shifts required for each insertion.