Abstract: Consider an arbitrary n-dimensional lattice Λ such that Z<sup>n</sup>⊂Λ⊂Q<sup>n. Such lattices are called {\it rational} and can always be obtained by adding m≤n rational vectors to Z<sup>n. {\it Defect } d(E,Λ) of the standard basis E of Z<sup>n (n unit vectors going in the directions of the coordinate axes) is defined as the smallest integer d such that certain (n−d) vectors from E together with some d vectors from the lattice Λ form a basis of Λ. Let ∣∣...∣∣ be L<sup>1-norm on Q<sup>n. Suppose that for each non-integer x∈Λ inequality $||x|| > 1$ holds. Then the unit octahedron $O<sup>n</sup> = \left{{ x} \in \mathbb{R}<sup>n:</sup> ||x|| \leqslant 1\right}$ is called admissible with respect to Λ and d(E,Λ) is also called defect of the octahedron O<sup>n with respect to E and is denoted as d(O<sup>n</sup>E,Λ). Let dn<sup>m</sup>=Λ∈A<em>mmaxd(O<sup>n</sup></em>E,Λ), where A<em>m is the set of all {\it rational} lattices that can be obtained by adding m rational vectors to Z<sup>n: Λ=⟨Z<sup>n,</sup>a1,…,am⟩</em>Z,a1,…,am∈Q<sup>n.</sup> In this article we show that there exists an absolute positive constant C such that for any $m < n $ dn<sup>m</sup>≤Clnmnnln(m+1)(lnln(mn)<sup>m</sup>)<sup>2</sup> This bound was also claimed in [1],[2], however the proof was incorrect. In this article along with giving correct proof we highlight substantial inaccuracies in those articles.