I'm not sure why EchoStar couldn't use a similar technique; unfortunately, it's possible that the courts interpret the patent broadly and figure that difference doesn't matter.
Such an interpretation seems probable. From the fine ruling, as linked from TFA (look for "2. Infringement of the '389 Patent"), emphasis mine:
EchoStar argues that because TiVo failed to prove by clear and convincing evidence that the redesigned devices infringe, it is not in contempt of the infringement portion of the injunction.
TiVo responds that the district could has previously construed the term "parses" broadly to mean "analyzes," a construction that has never been challenged, and all that is required for the modified software to meet this limitation is a component that analyzes video and audio data. TiVo argues that EchoStar's attempt to differentiate header data from the packet payload is improper because both are part of an MPEG packet, and the whole packet is video and audio data. Moreover, TiVo argues that experts for both sides testified at trial that PID filtering meets the parsing limitation under the court's claim construction.
The ruling is a great read, and great fodder for at least one aspect of what is wrong with software patents, even when they verge on hardware. EchoStar provides a DVR feature that is found to infringe a patent. In response to an injunction, EchoStar implements around the infringing idea by doing something it called "Indexless DVR". Now they're found in contempt of the injunction, essentially, because they found another way to implement the same feature.
This patent is, in effect, not being interpreted as covering an implementation. Or rather, the implementation being covered is so effing broad ("analyze this data and provide this functionality") that it may as well be covering something more broad than a single implementation.
This business of the container format being interpreted as part of A/V data is one thing that broadens the interpretation so.