Let k be the greates integer such that $2^k+k-2 \leq n$ (where also n is an integer). In this hypothesis $(2^k-1)/k$ is approximately equal to $n/\log n$. [For example it is always between $n/2\log n $ and $n/\log n$].
I cannot convince myself of the above statement. I tried some calculations but they didn't work out.
P.s: $\log$ is the log base 2.
An ugly approach. Clearly, $2^k \leq n$, so since $f$ defined on $[1,\infty)$ by $f(x)=(x-1)/\log_2(x)$ (and $f(1)=\ln 2$) is increasing, you get $$ \frac{2^k-1}{k} \leq \frac{n-1}{\log_2 n} < \frac{n}{\log_2 n} $$ (as long as $n\geq 2$ at least).
For the lower bound, by definition of $k=k(n)$ we have $ 2^{k+1}+k-1 > n $ and in particular, since $k \leq 2^{k+1}$, $$ 2\cdot 2^{k+1} > n $$ Using the same property of $f$, we get $$ \frac{n-1}{\log_2 n}<\frac{2^{k+2}-1}{k+2} $$ which implies (as $n-1 \geq n/2$) $$ \frac{n}{2\log_2 n}<\frac{2^{k+2}-1}{k+2} <\frac{2^{k+2}-1}{k} = 4\cdot\frac{2^{k}-1/4}{k} \leq 8\cdot\frac{2^{k}-1}{k} $$ at the end using $x-1/4 \leq 2(x-1)$ for $x\geq 2$. That gives $$ \frac{n}{16\log_2 n} \leq \frac{2^{k}-1}{k} \leq \frac{n}{\log_2 n} $$