Every dyno reads higher than every other dyno, unless it's your car - in which case the reverse is true. I don't care how accurate your dyno is or who made it, you will be told it reads high.
Seriously, no matter what you get, there will be bitching and internet experts.
If you're running other people's cars, a drum dyno is a lot easier to use. You can get cars on and off quickly. A hub dyno means a certain amount of screwing around to bolt it on.
I would STRONGLY recommend getting one from a big manufacturer. Dynos are like kit cars - there are a million out there, and mostly are powered by hopes and dreams. Also, expect to spend a bunch of time screwing around with it trying to chase down weird power spikes and the rest.
Dyno history at FM:
Dynojet (one big roller). Easy to use, consistent, had trouble picking up engine speeds sometimes. Couldn't do steady state tuning, sweeps only. The mass of the drum was the resistance. The four post lift to get the car on the dyno was actually pretty handy at other times. Sold it, it's still working at someone else's shop.
Rototest. High-end scandinavian unit that had all sorts of abilities. Used hydraulic pumps for load and strain gauges to measure torque directly. Very accurate and quick, pulled engine speed from the wheel speed and gearing. Could run for hours at a steady state, great for testing cooling systems. Had some odd behaviour when trying to do sweeps on big turbo cars due to closed-loop load control. Very finicky from a computer standpoint, any power loss meant it was difficult to get back up and running again despite the on-board batteries. It eventually died when one of the embedded NT boxes refused to boot up. Took forever to hook up a car. Didn't even have its own software for comparing runs, I had to write my own.
Dynocom: Drum dyno, very quick to hook up. Grooved drums means it makes a ridiculous racket and tears up tires. Has limited eddy brake capabilities for resistance, but you can't overwork it. Has had constant problems that "have never been seen before" and are constantly being blamed on installation despite the fact that we've done everything they say and more. Currently requires quite a bit of hand-holding from the operator to make sure the runs are even partially representative and we can't do steady-state. Cannot recommend.
I miss the Dynojet. Everyone who doesn't have one rags on them, but they're the standard for a reason. The Rototest was great for testing cooling systems and anything that required steady state, but it was a princess.
Customers will expect sweeps. The only complaint I have about the fixed speed tests is that they're unrealistically high load, especially on high power cars.