Problem about Algorithm 1 in paper

[ayachi3]

In Algorithm 1 , limits is sorted by the first index in descending order, which means limits[0][0] is the largest among limits[j][0].
So, either $s$ is smaller than limits[0][0], and be assigned to device 0, or $s$ is larger than limits[0][0], which means $s$ is larger than any other limits[j][0], and be assigned to device M-1.
In a word, only device 0 and M-1 will be assign to, what the loop structure for?

[Kuan-ChiehHsu]

The value s is a score output from the sampling_module(), which can vary depending on the sampling design. To determine which queue s should be assigned to, you need to perform a comparison based on predefined limits.

This isn't the assumption.

either 
s
 is smaller than limits[0][0], and be assigned to device 0, or 
s
 is larger than limits[0][0], 

[ayachi3]

Sorry, I still can't understand why that isn't the assumption. I understand that s can vary depending on the sampling design. But predefined limits is sorted by the first index in descending order, so limits[0][0] will be the largest.
For example, limits = [[10,0],[8,1],[5,2],[2,3]]
If $s < 10$ then $s < limits_0 [0]$ will be true. So $Q_i \leftarrow limits_0 [0]$
else, for each $limits_j [0]$, $s < limits_j [0]$ will be false, so $Q_i$ will remain M-1
As a result, only 0 and M-1 will be assigned.
Please tell me in which case, other device will be assigned.
Thank you again.