I'd be interested to know why your experience is different from mine. The data stream is the same for everybody and the blockiness is inherent to the way the data is encoded. It's certainly theoretically possible to run a smoothing filter over the edges afterwards to hide the defects, I suppose.