If you study the timings of human players, you will notice that they almost never play mechanically exactly what the theoretical written values would be. Timings, pitches, dynamics, everything. Musical notation is a means of written communication about musical ideas from humans to humans. It's meant to be subjectively interpreted by a performer.
Exact triplet swing tends to sound awkward, and the faster the tempo is, the more you have to straighten out the swing.
Why it's the default - maybe they listened to the 66.6/33.3 swing and decided that it didn't sound realistic? I think all of the things in the playback dictionary have been adjusted so that it would correspond to how people would actually play it. Another example is the note length percentages for normal vs. tenuto - I think it's just based on trying out different things and adjusting to taste.
The meaning of the metric modulation marking is actually explained in the Sibelius manual: (from version 6.2. Sibelius Reference)
I cannot provide any statistics, but in my experience the symbolic metric modulation marking is more widely understood than the English word "swing". And "swing 16ths" or "swing 8ths" would be even less widely understood... The symbols are a way to communicate the idea of swing without using English, even though there's the possibility of misunderstanding it to mean "to be played as triplet swing with machine-like mechanical microscopic accuracy".
To remedy the situation and cater for the microscopic interpretation - which might be becoming more and more common now that people experience life and communicate with each other through computerized filters instead of directly - it might be possible to use the "almost equal to" sign instead:
The downside of starting to use this would be, it would be contrasted with the old symbol, and it would enforce the incorrect interpretation of the old symbol as meaning "robotically precise triplets". Which usually isn't the original intention.
Perhaps the better alternative would be to educate the microscopic precision folks about the actual real-life use of the metric modulation marking? :)
FWIW, the metric modulation swing marking in Sibelius 6.2 seems to produce a swing of about 60% when exported to a MIDI file. Here it is imported to Ableton Live, and the background is set to triplet grid.
The swing isn't 66.6%, the length of the downbeat notes isn't precisely 100.0%, the notes don't have the same velocity...? That's because they specifically tried to program the Sibelius application so that its MIDI output would be close to a human player's interpretation of the notation.