The discussion may have moved on, but I'd like to review a bit what is known and what not really.
Some of the thoughts were centred around circling east and about an mistakenly chosen navigation source, mostly based on the debris trail.
The debris trail orientation is deduced from 1) the pictures of the crash site (
avherald, post #8 mirrors the second) and 2) from the map view satellite images. Both sources are spatially referenced by the creek running down the hill and steps in the terrain. While there is no doubt the trail is generally running in a north-south direction, the uncertainty of the exact direction is considerable, owing to the individual errors:
a) Uncertain point of view: the pictures of the site are taken from further away, likely from the airport, the second appearing to be taken from a more southerly position. The second picture shows also small dots arranged left to right (RWY?), towards the depression that is probably the northern end of the lake. That would, if correct, position the camera approximately at the latitude of the apron. Uncertainty is on the order of hundreds of meters.
b) Zoom skew: picture one is apparently taken with a strong zoom factor. Any variation in depth (along line-of-sight component) would appear smaller than real, so that an oblique line would appear much more parallel to the image plane. There are some spectacular near-frontal shots of cross-wind landings that demonstrate this, where the aircraft appears only to crab as much as 45° to the runway. Effective uncertainty maybe 10 to 20°.
c) Definition of the "trail": the north end of the debris field appears wider than the south end. There are larger parts in front but the smaller parts stretch further to the W. If one chooses the bigger parts and the tail to define a line, it would slightly climb hillwards and move away from the camera towards the south (cf. b) ). One may as well prefer a triangle or a curved southern end. Uncertainty maybe another 15°.
d) Georeferencing: the creek and accentuated steps in the topography (likely geological bedding) are the visual references to link source 1 and 2. The first picture shows a small blue object where the northern of the two elongated snowfield on the satellite map would approximately be along a prominent step. The location of the impact site is quite well constrained, but the step retreats toward the creek. That makes it more difficult to map the trail.
All in all, if anybody can determine a trail direction more accurately than say ±20°, then he is damn good (as a disclaimer, asked to draw a line, I would certainly do something similar like aterpstr on his map, maybe even tracking 300°). It seems mostly settled now, but the argument whether the track was parallel to the RWY or not, at least me, I simply couldn't answer, even less whether the last heading in the air before a (possibly oblique) impact was 347T or rather something else. (Even though SLF, I work in a profession where visual cues are important).
Similarly little is known about time, with an final call reported at 5 or 3 NM and apparently ~10 (±?) minutes before the fatal impact. None of this is reliable yet. Funny enough, the second most secure information after the location of the crash site is the general pre-impact attitude which was not extremely unusual, according to the recollection of a survivor which should at least be solid enough wrt extreme deviations.
Not to come with a grumbling "let's wait for more official facts", but being aware of how clues could possible fail in an assessment is always a step ahead.