The Robur does grinding and dosing very well but it's not perfect. Any system that associates Time to Weight will have some wandering associated. Coffee beans are super active water sponges. That is to say they are incredibly hydroscopic and ANY change in ambient temp and atmospheric humidity will change how fast the burrs will grind through the bean. The fill level of the hopper will make a slight difference, too. And the electric charge of the coffee in the window and the doser chute will effect how much coffee the window and chute hold onto. Mazzer has taken steps to MINIMIZE these things but they cannot ELIMINATE them altogether. Coffee is static sensitive for the same reason that it is water sensitive and even if it passes through a static screen when leaving the grinder throat it can pick up static charge very easily afterwards. There is a lot of electricity in a Robur, or any grinder for that matter, even the body will have tiny electric charges because it is coupled with Ground. Ground voltage DOES fluctuate. Not much and not for long but it doesn't take that much electricity to put a static charge into tiny coffee particles. They LIKE to have a static charge. So it would be logical to start to see some clumping in the doser chute, given enough time and enough coffee having fallen/moved/slid, (read FRICTION), through the chamber. Again, it's probably not much, and all the factors taken together probably account for some of this wandering. My experience is that just a few beans, 2 or 3, can be a gram of coffee, and when you think of all the places 2 or 3 beans can get hung up in a grinder it can add up pretty quick. The only way to get EXACTLY consistent gram throws is to weigh the coffee before it goes into the hopper and grind out the whole thing. Which is not a practical solution.
Some other questions to consider, when you say (+/- 4-5 g) do you mean that sometimes with a target throw of 21g you can get 16g-17g on the low side and other times you get 25g-27g, which is how I would read your statement, or do you mean that the total variation is 4-5g from lowest sample to highest sample? The way you have written the variance it looks like you are getting between 8g-10g difference from one shot to the next. That is a lot. And a bit alarming. On the other hand if you are getting a total variance of 4g-5g, for instance if your target is 21g and sometimes you get 18.5g and others you 23.5g, then that would actually be (+/- 2.5g) and very much closer to what I would expect from a timer based dosing system. Even with the Robur's many upgrades and well thought out design. Analog systems are approximate and messy. All engineers can really do is try to make the area of approximation smaller and smaller.
Second, have you checked out these claims yourself? The only thing you can really do is to pick a time when you are on bar, and when it is slow enough to weigh every single shot that comes out of the grinder, and then weigh every single one, noting how full the hopper was, how long between shots, and make a graph. You might be surprised to find that what is actually a relatively small variance gets communicated up the chain as an increasingly large one. You might also discover that the variance is associated with some external factor, like the sun coming up and directly striking the bean hoppers, or right when it starts to get REALLY busy, or when a particular barista comes on shift, or when you guys prop your door open for the morning rush or open the back to door to cool the place off. All sorts of things can factor into a situation like this.
I'd love to hear what your graph might tell you. Real numbers and real measurements don't lie. Especially if you are the person who did them. Just remember to use the same basket every time, clean it the same every time, zero out the scale every time and use the same coffee every time from the same grinder. Weigh the coffee before you tamp but you should also note the shot time associated the shot. There might be some interesting info there. And you will need at least 20 measurements to get any real info out of it. Probably more. 40 seems scientifically and statistically reasonable. The fact that you can get variance from one shot to the next is trivial. Good sleuthing and let us know what you find!
I should also mention that the Grind adjustment itself will make REALLY big differences in how much coffee the machine can grind in a given time. Theoretically, EVERY time you make a grind adjustment you will have to make a programing adjustment to the doser timer. That's just the physics of it.
hi Jeremiah, I've had similar problems with the robur e and was getting about 4g variance (+/- 2g) from dose to dose (especially with beans at different age). at the end I went back to volumetric dosing - using the basket as a guide where my variance is usually only about 1g (+/- 0.5g) . Not perfect, a bit more wasteful and messy but still better than I did with the robur. Really, pulling a 16g shot will be very different than a 20g shot when you dialled the grinder for an 18g dose (could some times mean up to 8 sec difference in shot time!), so it was worth it for me.