If you can post a sample packaged workbook (.twbx) example, I'd be happy to look over your work.  Its very difficult to know whats going on without seeing the actual data/workbook.

The reason you are getting this is that the % values have been rounded to two decimal places.  The raw % values are actually different (like 9.4281.... and 9.8096....) and when the raw values are divided, the result is what you are seeing.

Here's a version where I've explicitly rounded your percentage calc--I also changed your calculation to be:

SUM(i)/SUM(s)

Which is what you really wanted, I believe.  WIth the % field explicitly rounded to 2 decimal places (actually 4, since we're dealing with a decimal/percentage value here), the calculation will result in 4.03% as you expected.  Example attached.

But you were getting the correct results before--4.05% is actually the accurate number based on the raw values in your data.

if you have row level data, sum(i)/sum(s) is same as [i]/[s]. But when you aggregate your data, sum(i)/sum(s) will first count sum(i) and sum(s) and after that divide sum(i) with sum(s). So you will get weighted average.

If you use  [i]/[s] and aggregate your data, it will just sum those [i]/[s] together from each row. If you select avg instead of sum, you will get average of averages.