I'm assuming you're using different sized props to vary the load and that all readings are taken at full throttle with the optimum needle setting?

The curves don't look right to me either because even at the low revs you took your first reading they're both past what would seem to be their peaks. Even a mildly timed small 4 stroke would be giving a rise in HP (and most likely torque as well) from the initial (7000?) revs.

As for which direction to go when taking figures, it can be done either way but usually it'd be better to start with a high load/low rev situation because then you can see when the HP peak is reached and it's not necessary to go much beyond that.