If it's the PPM, I added a home-made baffle made from a prescription-bottle lid:
http://www.thehighroad.org/showpost.php?p=9433608&postcount=11
It now meters 4064 extremely well. It doesn't make much of a difference on ball or flake powders but it's a huge improvement on extruded powders. Ball powders meter fine, with a bit of leakage depending on the grain size. Flake powders don't meter very well at all.
Regardless of all of that, the key is consistency in the throwing process. When I lift the handle after throwing, I do it fairly sharply. Also, if your powder measure shares the same bench as your other reloading equipment, especially presses, their usage can cause powder to settle from transmitted vibration. So if you're loading process includes something like resizing 5 cases, charging them, putting them in a loading tray, and then sizing more cases, etc..., the first throw after resizing will probably be on the heavy side and the last one will be on the light side.
Just for fun, have you tried throwing 3 separate 30-grain loads and seeing if that creates a more consistent 90-grain load? I know several people who break up any load over 60 grains into 2 or more individual throws from the measure.
My opinion, and it seems to be shared by others here, is that 1 grain +- in a 90 grain charge is probably not that big if a deal, especially with extruded powders. If it were a ball powder, I'd be a little more concerned. If you want more precision, throw slightly light and trickle up to the charge you want. That's what I do when working up a load so I know exactly what I'm dealing with. Once I've found a load that works well, I'll usually load right out of the measure and weigh every 5 or so to make sure it hasn't drifted.
Side question: What type of scale are you using? If it's an electronic scale, skip the middleman and zero the scale with the empty case and use that as the "pan". The pan shuffle that you're doing right now isn't helping with consistency regardless of whether you're doing it correctly.
Matt