No CDI deflection when intercepting approach WPT

1. Load ILS26 via ZKF (FAF, so full procedure). Localizer frequency 110.7 tuned and ID'd. Set Direct To HAMES, the WP before the FAF (skipping the PT). Fly a course to HAMES that intercepts the inbound course at, say, 70 degree angle. IFD540 is showing GPS->VLOC. As I approach HAMES, the Course pointer and CDI are showing centered on course, then swing around to 255 (inbound track). 540 shows VLOC. I haven't yet started the turn, and haven't reached the inbound track yet, but the CDI shows centered, even though I am not on the inbound track yet. I can see the inbound course leg (from HAMES to FAF) parallel to the Course pointer. I think that the CDI should be showing a deflection in the direction of that course leg. Instead, I ignore the CDI indication and fly to the depicted inbound course track. Everything works fine after that.

2. Essentially the same description, but on an RNAV approach, for example RNAV 08. Set Direct To LESAV (IAWP). Fly a moderate angle towards LESAV. The Course pointer is pointing to LESAV, CDI is centered. A couple of miles away, the Course pointer swings to the next waypoint (DUPUR). Even though I'm still a couple of miles away and not on course to DUPUR, the CDI is centered. I can see the next leg depicted parallel to the Course pointer, so I fly that to intercept the actual track. I am expecting the CDI to show the lateral deflection.

In both cases, following the course pointer would result in turning too early and too fast, since the pointer swings and centers immediately, even though I'm not on the new course.

Am I missing something obvious? If so, I haven't been able to figure it out. If this is discussed elsewhere, pointer will be appreciated.

The Aspen image shows the issue. I am flying a track 112 degrees off from the first leg of the RNAV 05 approach into CYFD, waypoint MIBLA. In the image, the course pointer has just swung to show the desired track from MIBLA to OVUTU. The magenta line continues to MIBLA, the white line is the track to OVUTU. The course deviation is zero. The IFD540 cross track distance and deviation are also zero and DTK is 142, so the Aspen and 540 are in agreement. To me, that takes the Aspen out of the picture. The behavior is being driven by the 540. (That is good news from a troubleshooting point of view.)

I verified the 540's GPS behavior on the IFD simulator. It's the same using both the old sim (for 10.1) and new sim (for 10.2). When approaching the waypoint, the DTK changes to the course for the next leg, the deviation is zero.

What I have determined is that, as long as I am on the curving GPS track that provides the turn anticipation, the deviation is zero. That may be logical since I am "on course", but it disagrees with the DTK.

I asked my buddy with a G1000 to try this. At a similar point approaching the IAWP, the course pointer/DTK change to the next leg, but the deviation is NOT zero, it is showing the actual deviation from current location to the track of the next leg. I think that is correct behavior. At least it is more logical to me.

I find the 540's display of zero deviation to be non-intuitive. I suppose I can learn to live with it, but I'm looking for either a way to change the behavior, or an explanation of why this is better or correct behavior. It doesn't look that way to me.

I still need to try this in the air with a localizer. I know that, when I am flying procedure turns and returning to the localizer, the deviation works as I expect, guiding me to the localizer. Yet, when I am not doing a procedure turn and just intercepting the localizer, the deviation works differently. That has been confusing. I need to verify it again, and capture some images. I am expecting that the zero-deviation behavior will occur while in GPS mode, but the actual deviation will appear when the 540 changes to VLOC. I want to try making the change to VLOC manually. That would provide a workaround for intercepting localizers, but I'd still be faced with this unwanted behavior in GPS mode. To me, it should always work the same way.