I'd agree with Chalmers (post #2 above) that a poll would be interesting.  I would guess for example that Derek falls into category #1 and Stevan falls into category #3. 

The suggestion as to why we should feel at all is often used to aid in getting across the concept of an explanatory gap.  However, the following might also have been suggested in the literature - though I've not seen it personally and would be interested in feedback.

One can either claim phenomenal consciousness is epiphenomenal or not.  Certainly, a transistor for example changes state because of a voltage/current applied to the base.  It will never change state otherwise making downward causation impossible and p-consciousness epiphenomenal as Stevan Harnad suggests.*  We might make a similar claim of neurons though I'm aware of at least 2 reasonable objections** to this perspective in the literature and there may be others. 

If consciousness is epiphenomenal, we might claim the experience reliably corresponds to behavior as Stevan suggests.  Assuming it is epiphenomenal however, means that the phenomenon can have absolutely no influence whatsoever on behavior.  The experience can be anything, meaning it can be pain, color, auditory experience, or any category of experience as suggested for example by Chalmers (pg 6, "The Conscious Mind").  It could also be any experience NOT categorized, ones we have no subjective knowledge of whatsoever!  Note also that if it is epiphenomenal, we can't assume the phenomenon will be reported reliably.

In other words, forget about why we should experience anything at all.  If p-consciousness is epiphenomenal, we might ask for example:  Why shouldn't I experience only a whistling noise, a photism, or just pure euphoria as I drive to work, stopping at intersections and avoiding other traffic?  I could experience anything but I would still tell you that I see the car approaching in the distance and I'd still report all spatial and auditory information such as distances to oncoming traffic and sounds from horns and broken mufflers exactly as if my phenomenal experience reliably corresponded to the actual, physical information my brain was able to receive through visual and auditory sensors.  But I wouldn't experience any of that.  I might only experience something utterly uncorrelated to the actual environment.

If p-consciousness is epiphenomenal, the explanatory gap can be viewed from the perspective of, "Why should the experience produced correspond to reality instead of simply being a phenomenal experience which has no correlation whatsoever?"  If p-consciousness can't influence behavior, then this phenomenon can not make itself known by adjusting behavior nor reporting the actual phenomenon that occurs upon penalty of telekenisis as Stevan points out.

In addition to providing personal opinions, I'd be very interested if you would be so kind as to suggest papers or literature that might address this perspective.

*I'm sure there are those who would disagree with this and allow for some kind of downward causation in a conventional computer system, but it really doesn't matter.  The point regards epiphenomenal p-consciousness.

**The 2 I'm aware of:  1. The argument that nonlinear physical systems are in some way holistic/non separable (esp. Alwyn Scott) and therefore capable of downward causation and 2. quantum mechanical systems.