Created attachment 606 [details]
Spreadsheet showing the problem

Currently Inkscape's rendering of filters-gauss-01-b.svg is different from the reference PNG. However, this is only true for the regions where two rectangles overlap. This is most apparent with the blurxy test. The top and bottom parts match almost perfectly while the middle part is clearly different.

I have attached a spreadsheet that shows some graphs demonstrating the problem. (For the reference PNG only the green channel is shown to reduce unnecessary clutter.) The graphs correspond to lines extracted at y=220, y=270 and y=320 (from top to bottom). Legend:
  - the continuous green line is Inkscape's output
  - the dashed green line is the W3C reference PNG
  - the brown line is the result of blurring a 45 pixels wide strectch of
    "red" roughly corresponding to the visible part of the red rectangle
    (only the green channel, so all 255 except for the "red" part, which
    is 0)
  - the orange line is the result from alpha compositing the green
    channels from y=220 and y=320 (corresponding to only red and only
    yellow).
As can be seen from the middle graph Inkscape's output matches quite well to the data computed in the spreadsheet, while the reference PNG's line differs quite a bit. The orange line was included to rule out the possibility that the difference was caused by filtering the rectangles separately vs. filtering the combination, apparently it doesn't matter much which approach is used (for this method).

So, which image is correct? (And what causes this mismatch?)