The real question is... do the darker steps LOOK brown (you get yellow or brown when you combine red and green light... yellows when there is a lot of light, shifting to brown as the luminance gets darker). If the steps don't LOOK brown, you can't trust your meter.
Here's why... colorimeters are LIGHT meters, not DARK meters. Something has to be there for them to measure or they simply get inaccurate without TELLING you they are getting inaccurate.
Let's start with a projector that produces 15 fL measured off the screen for a 100% white test pattern. About 72% of that light is in the green spectrum. About 21% of that light is in the red spectrum, and only about 7% of that light is blue light. That's how human vision works... those (approximate) percentages are what our eye/brain interface interprets as neutral gray or neutral white (when the display is nominally d65 - which is a far more accurate way of stating that the color temperature of the display is 6500K... 6500K doesn't tell you much about whether there is too much or too little green light... d65 is the precise reference point we want the display calibrated for and it includes all 3 primary colors).
So with just 7% of the light being blue light, and with 15 fL for 100% white, that means you (and your meter) are "seeing" just 1.05 fL of blue light when you are displaying 100% white. 75% white is just about "50% stimulus" meaning that a 75% white pattern will be roughly 1/2 as bright as 100% white - or 7.5 fL. So at 75% white, blue is already down to (roughly) 0.5 foot-Lambert. That's getting into a range where the meter is ALREADY going to be struggling with ACCURATE measurements of blue light... simply because 1/2 a foot-Lambert isn't much light to measure. The less expensive the measuring device (a general statement), the less likely the device is to be good at measuring low light levels, primarily because there's so little blue light the meter just can't "see" it. In 2006-2007, a well known spectroradiometer selling in the $10,000-$15,000 range depending on options, was, according to company rep "pretty much done measuring accurately below 1 fL (of white or gray light)." Fortunately, those models have been replaced by newer, more capable models... but STILL, measuring levels at or below 10% (ACCURATELY!!!!) with a great meter selling for $6000 - $30,000 is difficult or impossible. You can't expect the same level of performance at darker steps from meters selling for $1000 or less. There are a lot of meters out there that are totally untrustworthy below about 20% white for a flat panel or below about 30% white for a projector. Remember, projectors are typically half or less than half as bright as a properly setup flat panel display!
If you were measuring a flat panel display that was setup for 35 fL for 100% white, your 75% white pattern would have 1.225 fL for blue light. You'd be able to continue to get credible readings probably down to 20% with a reasonable meter like the Colormunki. But you won't measure 20% white (or lower) accurately, particularly in the blue spectrum, simply because there's so little blue light in what we interpret at neutral gray (d65). Simply letting the meter read for longer periods of time doesn't typically help because you get into the noise floor of the electronics in the meter and the noise modulates readings to such a degree that the readings are still inaccurate, and may vary quite a lot if you were to take, say, 10 readings and try to average them... you end up just averaging the noise floor of the meter which is no help when you are trying to measure darker grayscale steps.
These meters never TELL you they are at their limit of accuracy. They continue to produce readings for quite some time before you finally get the dreaded "not enough light" error. Even the $14,000 meter I use does the same thing... below about 10% white for panel displays (about 20% white for projectors). The meter (and software) continue to produce readings down to even lower than 5% white. Somewhere below the 5% white luminance level, the meter will finally not produce a reading at all and the software will report "not enough light" or something like that. But the truth is, the 10% reading is probably starting to get inaccurate (for a panel display) and nothing below the 10% reading should be trusted (again for a panel display, we're probably at 20% with THIS meter for projectors and screens - though an ultra high-gain screen might produce a higher threshold. "Lesser" meters will have similar limitations, but typically at higher light levels simply because the noise floor of the instrument begins to equal the signal level of the light being measured. You end up with readings... they just don't mean much and nothing tells you they are "suspect" or "inaccurate". But they are.
