I suggest that the rationale may be somewhat simpler than CB's comment and rather similar to Centaurus' suggestions.
The discussion below is my own long held view as to the thinking behind approach additives - one would need someone from Boeing Aerodynamics/Flight Test/Certification to resolve the specific question with authority.
Like the boundary layer associated with flow around an aeroplane, a steady wind near the ground (assuming no obstacles) reduces as the aircraft approaches the flare. For flight test and certification purposes, this is normally modelled using the "one-seventh" power law - if your AFM has a wind graph which varies with height above the runway then this model is most likely what is built into the picture. The practical thickness of this "boundary layer" is, however, considerably less than 2000 feet.
By allowing a modest additive (half the steady headwind component) to the basic AFM Vref we try to provide for this wind component decay, considering that, on most occasions, we have no idea at what height the (tower) reported wind was measured.
The thinking is a bit like that used when talking about undershoot shear - one is trying to preserve a high probability of not finding the airspeed too low entering the flare. As the wind profile is reasonably predictable for a steady wind, and the tower wind height is not likely to be less than around 30 feet, it is appropriate that this speed additive be bled off as the aircraft enters the flare and touchdown. The whole thing is a bit waffly, so it is not critical to worry too much about whether we start the bleed reduction here or there - more important to hold a reasonable margin on until the aircraft is approaching the runway. Perhaps coming over the fence is a reasonable point to think about permitting a speed bleed.
For a gust, on the other hand, there is an element of randomness, so it is appropriate to carry the full gust value. Further, due to the randomness consideration, it is appropriate to maintain the gust additive through to touchdown.
To guard against touching down with too much speed for comfort (keeping in mind that a constant stopping effort landing roll has something to do with speed squared) it is appropriate to put a cap on how much of an additive one carries - Boeing opts for 20 knots maximum. Provided the pilot controls approach, speed, and float within reasonable margins, the normal 1.67 distance factor provides sufficient fat to allow for the modest increase associated with approach speed additives. In the case of wet runways, the normal distance penalties (typically 15 percent) provide an additional acceptable balance for speed additives.
This seems to sit well with the Boeing recommendation for the 737 (FCT 737 (TM) Chapter 1)
".. add one half of the tower reported steady headwind component plus the full gust increment above the steady wind to the reference speed .... The gust correction should be maintained to touchdown while the steady headwind component may be bled off as the airplane approaches touchdown"
Furthermore, as Mr Boeing thinks that his autothrottle is much better than you or I, the recommendation is that the wind additives be applied only for manual throttle operation. The autothrottle is presumed to be able to catch any wind variations.
Any other thoughts .... ?