When i used to dyno my dragbike it was on a rolling road and that worked in the usual way of calculating resistance on the run down to provide a flywheel (or crank) figure . The problem i had was sticky tyres in my case slicks, so the first run came out pretty high (in the opertaors opinion) and subsequent runs after the tyre got warm often sent the calculations into error and aborted the run losing any data. I took 18 runs to get a decent one out of the dyno and truth be told i wasnt interested in the figures just the afr and a nos test.
Fast forward to buying my Skyline and the car came with a dyno sheet but it stated on the sheet the drag output was 75.5hp which was very high imo for a rwd car after a few mods and the car was taken to the same dyno and the drag was 72.5hp
Apart from the changes to the engine i had actually fitted better tyres going from some crappy 235 dunlops to 255 toyo proxy's so i was expecting the drag to go up if anything.
My last dyno run was on a hub driven dyno and the graph in comparison is much smoother , unfortunatly the figures are lower so you do lose on the bragging rights

As a last comment i have a mate who owns a bike dyno and he did say he had had a few customers ask him to alter the parameters on the dyno to give inflated figures if they were selling the bike or simply wanted bragging rights
So as a tool to dianose faults and to establish an engines performance they are great but the figures should only be compared on the same car . Comparing different dynos figures is a waste of time IMO