Yes, that is pretty typical for these kinds of mounts.
You might look through the other message threads on this forum, there are several discussions of the Juwei 17 mount and some questionable mount performance.
In general you need to guide this really aggressively.
use 1 second exposures or less.
Consider using RA PPEC algorithm with a period of 422 seconds (based on your results) and disable auto adjust period. you will need to evaluate aggression settings for both reactive and predictive. There is a tradeoff between high aggression, exposure time, and possibility of introducing oscillations. You will need to find the right balance for your mount.
PHD can effectively compensate for significant amounts of error, but there are some errors too large to address. The Juwei 17 seems to be one of those with exceptionally high error. Another customer reported a PE of 60"