Hello Al
... Timo's XLProfiler to create your profiles?
The rendering intent shows up as perceptual
Photoshop often shows the Perceptual intent as the default setting even if the selected profile does not have that rendering mode at all, it simply reads the Relative intent conversion from the profile and says that it is Perceptual intent. Well, they are actually the same for input devices.
The inCamera application doesn't
seem to allow that ... only Relative
It is behaving correctly, XLP profiles have relative intent. XLP profiles can be created by Saturation or by Absolute colorimetry but inside the profile the conversion is always calculated and specified by Relative (to overcome some problems that some color-management engines have, most do the relative intent correctly).
They also don't allow a black-point setting
ICC profiles (none of them) do not allow nor deny blackpoint setting, that is up to the user at the post-processing session. The XLP profiler does not allow to clip the blackpoint (so there is no blackpoint adjustment in the XLP), this is the only proper way to write a profiler, however XLP does make sure that the dark end will reproduce accurately, there are other ways to do that than the crude blackpoint clipping.
how did you end up checking the resulting profiles?
When the created profile is Attaced to the target image that was used for creating the profile then if the on-screen appearance is an accurate representation of the actual target then the profile is very good.
There are however other profile quality issues, most importantly one has to evaluate if the profile is clipping the grayrange. This is very simply esst to do just by Attaching the profile to a gray test CGI that has the 256 gray pathces from R=G=B=0 to R=G=B=255 and then converting the data to a linear working-space, by inspecting with the eye-dropper in Photoshop one can easily see if the profile is clipping at either end of the range. XLP profiles do not clip but they are using 12-bit TRCs (where most of the commercial profilers only write 8-bit TRCs) so the eye-dropper that is also using the 8-bit gradation even if the data is in the 16-bit/c mode will give erroneus reading, therefore in order to see if there is separaration when using XLP profiles one has to use the Levels or Curves dialog to scale the converted image, once for the dark end and once for the light end.
been using ICC Inspector to view the CIE
chromaticity chart ... is there a better way?
The gamut of a tri-chromatic device is a 3 dimensional volume (somewhat like a 3 sided pyramid), not a 2 dimensional area so yes there are better ways, but such sw are rather expensive and I've only seen such sw for the Mac OS not for Windows.
the profile I make may be flawed and I wouldn't
really see the problem until a particular image
demonstrates the issue.
Without profiles each and every image you take is flawed. If a profile does not clip the dark nor the light end then the profile is good for all the images. One issue with profiles is the whitepoint, many people are using profiles that were created for Daylight (sunlight) and they post-process any WB correction if needed. The other way to overcome this is to create a lot of profiles for all the possible illumination conditions but that is a little bit too much of work considering that the WB correction is a very easy post-process operation.
The sizes of the profiles are much larger than
I expected. Is that because they are all LUT?
ICC profiles can have 1 dimensional LUTs and/or 3 dimemensional LUTS, the lattter is called cLUT (Color look up table) and the former are transfer curves (TRCs).
Profiles that use 3 dimension LUTs are usually very large say from 300kB to several megabytes, these are crude/coarse profiles even if they are extremely large because they actually do not convert, they are just conversion tables and rely heavily on interpolation both at the profile creation phase as well as when the profiles are used. The result of the interpolations is that the amount of the individual colors in the images are strongly reduced after the profile conversion, this can be easily checked by Assigning the profile to a test CGI that has e.g. the 16.7 million RGB colors of the 8-bit/c mode and then converting that to a linear working space and counting the remaining individual colors (can be done e.g. with the PaintShopPro, Photoshop does not have such a tool).
E.g. XLP writes about 25kB profiles, it has 12-bit/c TRCs on R, G and B channels followed by a 3x3 gamut matrix (that has 16-bit/c gradation). This kind of profile does specify the actual conversion, it does not stretch/tear the data by interpolation and it does not reduce the amount of color in the image in huge quantities after the profile conversion.
It appears exposure has more of an impact than I originally
figured. ... Did you find that an issue with Timo's application?
XLP does a lot of work in order to overcome the problems of under or over exposed target shots. With what ever profiler the profile quality always suffer in case the target shot is underexposed (histogram is not full), with many profilers the profile quality suffer also if the target shot was overexposed. With all the profilers perfect exposure (histogram is full) of the target shot always results in the best quality profiles.
XLP ignores overexposed and underexposed pathces of the target so it can end up with rather good profile in both cases (especially in the case of overexposure) but naturally if the histogram of the target shot is not full then that portion of the acquire device behavioural where we have no information at all can not be profiled well, not with any profiler (it can only be guesses). The profile structure that XLP is using however is much better in "quessing" the behavioral in that "no data" portion.
Timo Autiokari
http://www.aim-dtp.net 
