****FURTHER UPDATE****Several days ago we discovered the FD in each of our CV880 and CV990 panels is not working correctly in both "AP VOR/LOC" and "AP ILS" modes.
The FD "is" however working correctly in "AP HDG" mode .... only.
In the case of AP VOR/LOC mode, the FD indication remains pegged to the extreme left or right of the AI gauge as follows ....


Despite these FD indications so long as one's intercepted, and successfully acquired, any pre-selected VOR radial our CV880/CV990 simulations "WILL NOT" drift off-course and for the duration of any acquired radial .... unless one disengages AP VOR/LOC mode in attempt to manually adjust heading in order to try'n align an FD indication in relation to the intended VOR course.
In the case of AP ILS mode, the FD indication also remains pegged to the extreme left or right of the AI gauge also. So long as one's intercepted, and successfully acquired, any pre-selected ILS/GS radial our CV880/CV990 simulations "WILL NOT" drift off-course and for the duration of any AP ILS/GS coupled approach to landing either .... although disengaging the AP and VOR/LOC momentarily, then re-engaging both once again ("NOT RECOMMENDED" though in the interest of maintaining a stable coupled approach to landing) may then force a correct FD indication as follows ....

It seems these recently discovered FD related issues have been applicable to both panels since they were first released during 2003. Neither we, nor anybody else either, appear to have detected and reported these issues all this time/to date "despite the fact this issue is there to be seen" (none of our development team are pilots, so, it's something easy to miss) .... or if previously reported then it's not been clearly understood by us

I don't know if this FD related issue has any bearing on what "THALES COELHO" has experienced or been trying to communicate though.
Other than the above reported observations I can't find any major problem/s, at all, "in relation to the CV990 AP NAV features" (which incidentally are identical to those featured within our CV880 and DC-8 panels too) other than the following illustrated observation ....
After intercepting, capturing, and stabilizing upon any particular VOR radial. If one then selects SHIFT+Z (to display default FS LAT, LONG, HDG, ALT, SPD, and WIND etc data per that red text which appears within the top left corner of any main panel view), then, upon comparing the "HDG" indication displayed by this data with that of the intercepted/captured VOR course indication within the HSI gauge, one may note despite a perfectly aligned VOR NAV capture (in accordance with any selected course) the default FS "HDG" indication (again per that red text) is at variance with the HSI indication by some 2* or 3* degrees .... as is evident within the following image ....

In the above image the simulation has successfully intercepted, captured, and is aligned with the selected 156* VOR radial/course .... upon which it's been well established, in the cruise, for some 159 miles DME .... but .... the default FS information (again that red text) indicates a course/HDG of 159*. This 3* variance "IS NOT" the result of the simulation very gently/very slowly weaving fractionally left and right over the captured VOR course .... and as is normal within FS anyway. "NOR IS IT" the influence of winds aloft either (I checked that too). During every test I conducted this minor discrepancy remained "CONSTANT" and throughout the entire 194/5 DME duration of any selected/captured VOR course. This's only evident when using AP VOR/LOC mode though .... and regardless of the route and course being flown. The interesting thing is the simulation "DID NOT EVER" miss or stray markedly from its intended destination VOR. Any attempt to try'n manually correct such an observed discrepancy (in relation to the displayed default FS HDG data at least .... not in regard to the HSI indication) could encourage the simulation to slowly drift off course/away from the intended VOR radial though.
As part of my checks, and whilst observing this minor discrepancy, I also engaged AP HDG mode whilst flying the very same VOR radial/course. In doing so both the indicated HSI VOR course, and the default FS course HDG information (that bloody red text again

) were this time in agreement .... as per the following image ....

In all aspects of navigation there "IS" such a thing as natural deviation, but, so far as I'm aware FS doesn't replicate this.
Therefore my final advice (to everyone flyig our CV880 and CV990 panels) is simply as follows ....
Set a VOR radial/course in the HSI gauge.
Set the relative NAV1/2 radio frequencies.
Use AP HDG mode to intercept the selected VOR radial/course .... from no more than an axis of approximately 25* degrees (for the quickest and most stable VOR capture).
Engage AP VOR/LOC mode in order to capture and maintain and maintain the selected/intercepted radial/course .... and just fly normally.
Alternatively engage AP HDG mode in order to pursue a course between 2 selected VOR radials .... and just fly normally.
In the case of an AP VOR/LOC intercept "DO NOT" rely upon the FD indication within the AI gauge.
Using AP VOR/LOC or AP HDG modes one won't/shouldn't be led astray" .... unless one opts to disengage AP VOR/LOC mode and adjust their course manually in attempt to seek an FD alignment.
Again .... the FD indication is functioning correctly in AP HDG mode though.
ONE LAST THING ABOUT ALL THIS .... 
I did try'n to correct the above illustrated and commented discrepancy by importing different variations of AP NAV data within the PANEL.CFG (such can often be done in order to influence/correct gauge indications). In doing so I "WAS" able to correct the FS data indications "by 1 degree only" .... barely any gain at all for the applied effort, so, I simply reinstated the original AP NAV data values.
This particular issue, and that in relation to the FD indication too (in both AP VOR/LOC and ILS modes) is likely within the gauge itself (possibly), but, as we didn't compile that particular gauge we can't amend it .... and I can't provide any indication in regard to when we might be able to resolve it .... if at all.
Again I don't know if this FD related issue, and minor VOR versus HDG related discrepancy, have any bearing on what "THALES COELHO" has experienced or been trying to communicate.
Mark C
AKL/NZ