The real lesson is not so much that renderers or controllers can be of different "quality" but that you never know what a renderer is going to do. You would expect a renderer to get the file, and then just play it without modification. But you can't really be sure that's what it's doing, unless you try and measure it.
It sounds like the TV is doing some sort processing of the audio. It might decode the DLNA file, process it, and then send it out the optical port. It might be doing SRS, bass/treble, or who knows what. But it seems it processes DLNA data differently than HDMI, which perhaps it is just passing through (as it should).
If you have the means (REW and a calibrated microphone) then you could measure the output and actually get a very good idea of what's happening.
Regarding the Sony vs KEF wireless: even a 0.5 or 1db volume difference can color your assessment of which sounds "better". Assuming the Sony is passing the PCM data unaltered, the should sound identical, so again measuring would allow you to exactly match the output levels to assess that.
As an aside, you said you had three renderers, but only mentioned two (Samsung and Sony). If you are playing DLNA wirelessly to the KEF, the speaker itself would be the renderer, and MC would be the controller.