If there exists a positive integer p≤⌊n/6⌋ such that p∤n, then we have ⌊n/2⌋>⌊n/6⌋, and taking k=⌊n/2⌋−p≥2 and two positive divisors d,d+k of n, we need d+(⌊n/2⌋−p) to divide n. But d+(⌊n/2⌋−p)≥d+⌊n/2⌋−⌊n/6⌋>d+(n/2−1)−n/6≥n/3, so d+(⌊n/2⌋−p)∈{n/2,n}, the only possible divisors of n larger than n/3. However, d+(⌊n/2⌋−p)=n/2 yields d=p, absurd (since d∣n but p∤n), while d+(⌊n/2⌋−p)=n yields d>n/2, thus d=n (since d∣n), forcing p=⌊n/2⌋>⌊n/6⌋, again absurd. Therefore all positive integers not larger than ⌊n/6⌋ must divide n.
Denote u=⌊n/6⌋. Since gcd(u,u−1)=1, it follows u(u−1)∣n, so u(u−1)≤n=6(n/6)1 cannot be odd (which is trivial), and by the fact that the only k defect for n=36 is k=13, the idea for the solution given above. Anyway, a fresh idea, and which is implemented elegantly and with minimal calculations.