Why don't the values given by Argyll's scanin match what I see in darktable or Gimp

I’m trying to do a variation of the color profiling flow given in this post about color profiling

I take an image of a Wolf Faust it8 target, turn off a bunch of the default processing that darktable does, change the input, output and internal profiles to linear rec709 RGB and export to tif.

When I use the eye dropper tool in darktable on the top right patch (A1) I get RGB of 59,51,51.
When I load the exported tif file into gimp, I get basically the same values. Great.

When I run Argyll scanin using the -o flag (which generates a list of measured values):

scanin -v -dipn -o IMG_1956it8.tif it8.cht

I get

SAMPLE_ID RGB_R RGB_G RGB_B
A01 11.60257 8.594904 8.657046

totally different values. Why?

If I change the exported profile to srgb, the values are ballpark but still off by 10 or more (so unusable).

I understand why the srgb values off (At least I think I do), but why are the linear rec profiled values so wrong? It feels like scanin is ignoring or is confused by the profile.

As a last data point, when I look at the tif in feh, it looks way dark.

My files can be found here. Sorry it’s 60MB but it includes my input RAW/CR2 file.

Thanks for reading. 🙂