The dyno measures tq, and calculates the hp.

Since all the results are above 5252rpm, the differences from the tq curve will be exaggerated in the HP curve.

My guess is, looking at the jog in the tq curve, it was some sort of brief misfire(drop in tq), then the load cell gets a little tug when the misfire recovers(spike in tq).
If you applied some “smoothing” to the curve, in that area of the curve, the numbers should be more representative of reality.
Unless of course you see the same jog in the data on every run, in the same place…….. which still wouldn’t rule out a misfire, but would indicate something is creating the misfire(if that’s in fact the problem), and that it’s not completely random.