The problem is that for the paper you are using, the L* of a heavy yellow inked patch is higher than the L* of an completely un-inked patch. RGB 255,255,255=L70, RGB 255,255,0=L75. While solid yellow patches have high L* relative to white on M0/2, they are always lower.
This is an artifact of the M3 process. un-inked areas reflect a lot of polarized light on metallic papers which is removed by M3. OTOH, yellow ink converts polarized light to unpolarized light and, while it reduces luminance the least of any color, the reduction in polarized light is greater leaving un-inked patches measuring lower L* than solid yellow ones.
Anyway, Argyll can make a profile though it exhibits an odd increase in area when near the white point which reflects the odd behavior on strong yellows.
Attached is the gamut slice at L*=69 and the Argyll profile in a zip file.
The white line is the gamut of a typical Pro1000 glossy profile.
Update:
Looking at the Argyll profile's internals, it's clear that the Argyll s/w is using the unprinted (RGB=255,255,255) as the actual paper white point and simply clamping L* when other measured patches exceeds that luminance. Presumably (reasonably) assuming the high values are in error. So I did the same thing to the measurement file and here is a list of before and clamped L*. They all, as expected, occurred on the yellow axis.
255 255 0 75.46 -> 70.06
255 255 28 72.06 -> 70.06
255 255 56 71.40 -> 70.06
255 255 85 70.34 -> 70.06
255 255 113 70.40 -> 70.06
255 255 141 70.36 -> 70.06
255 255 226 70.12 -> 70.06
Submitting the clamped CGATs file to I1Profiler generated a profile which I have attached in addition to the Argyll profile.