Sample size determination for finding the "top 100" largest X among multiple datasets My data sources

*

*Source A gives me a ranking of the top 1.000 largest X (say, companies).

*Source B provides a similar ranking, comprising 7.000 X, with some differences to Source A (different values, different companies).

*Source C has another list of the 5.000 largest X, again with discrepancies from A and B.

The result I want to achieve
I want to combine the three sources (A, B, C) together to generate a new ranking, one that is supposed to give the top 100 of the largest X.
I would manually determine the value for each X. (For example, one of the X -- called "XYZ" -- may have a value of 300 in source A, a value of 280 in source B, and a value of 279 in source C, but I would manually try to determine the "true" value of XYZ, which may actually turn out to be 400.)
This new list is supposed to be "better" because it corrects for each source's omissions.
The question: Sample size?
If I am content with generating a "top 100"-list (based on a combination of A, B and C), how large should my sample be (for manually determining the "true" values)?
Suppose that the the three lists (A, B, C) have 10.000 unique companies in total (unique X), but I do not have the resources to manually look up the values for all of them. As I am only interested in the final "top 100", can I just reduce my efforts to look at the "top 300" of each list (A, B, C) and then determine the final "top 100"? (Premising that there is nothing below the 300rd rank whose "true" value would propel it to the final top 100-list.)
 A: It depends on the ranking you use for your combined list. If your ranking e.g. says that, to be among the top 100 in the combined list, you have to be, as a condition, among the first 100 of all of the original lists, then you save yourself a lot of work because it is sufficient to only consider those in the top 100 of all the source lists.
Now, let's say you use as ranking the sum of the ranks for all the original lists, where the company at the top of list A gets the rank 1 for this list (and you set the rank for a company that is not appearing in a list of 10,000 to 10,001 for this list). For instance, if a company F is at the top of list A and list B, but doesn't figure in list C, you give it a rank sum of 10,003. Then the worst case for a company that is at the top of list A would be to get a rank of 20,003. That is, there are a lot of possibilities how a company could get a better (i.e. smaller) rank, e.g. one that is at rank 1000 in each of the three original lists.
So, in this case, I would proceed as follows: check the top companies in the three lists round robin and assign as total rank their rank sum. Also, you count your steps, because this count is a lower limit of what rank can be achieved. E.g.: Let's say you compute the rank sum of the three companies at the top of the three lists. After that, you record the count $c=3$. And clearly, at this point, no company that has not yet been checked can get a rank sum better than (i.e. below) $c=3$. You continue like this until you have 100 companies with a rank sum below or equal to $c$.
