As with a lot of things in the hobby, what seems like a simple question can prove to be otherwise, and it's difficult to generalize an answer. In the end, it partly depends on the way in which PEC is implemented for your mount and people, by nature, tend to offer advice based on the particular mount they have. I would say there are 3 scenarios that can occur:
1. PEC is implemented in the mount firmware and the model-driven RA tracking corrections occur predictively and don't conflict with external guide commands. This is the best-case scenario because the corrections occur before PHD2 will ever "see" the tracking errors - PHD2 just sees a mount that performs better.
2. PEC is implemented in the mount driver, and that driver is capable of processing both model-driven corrections and PHD2 guide commands in a sensible way. This isn't as effective as scenario 1 but it is still workable.
3. The mount driver or the mount firmware has a flawed implementation of PEC in the sense that it doesn't intelligently handle both model-driven and external guide commands. It may throw up its hands and do nothing, take action on only one or the other, etc.
Unfortunately, you may need to learn which of these scenarios applies to your mount. For all scenarios, the PEC model needs to be accurate and based on a training data set that excludes guide star deflections caused by things other than native RA tracking errors. A bad PEC model is likely to produce bad results in all cases.
Mount PEC models nearly always confine themselves to dealing with tracking errors having periods that are harmonics of the mount worm period. In other words, they can deal with errors having periods of w, w/2, w/3... where "w" is the worm period. They can't generally do anything about errors that have periods like 2w/3. This isn't true of PPEC in PHD2 because PHD2 doesn't know the native worm period and doesn't need to use it for the model it builds. So it is capable of dealing with a tracking error having a non-harmonic period - which is another reason that PPEC running on top of an accurate mount-based PEC can produce very good results. As a final note, higher frequency tracking errors create more challenges for guiding while a long-period error, even if large, can often be "guided out" fairly easily. But as the period of the error decreases, it becomes harder for the guiding software to fully correct for it. That's why it's often a good idea to "lock" PPEC on a lower-period tracking error even if the peak-peak magnitude of that error is smaller than other errors having longer periods.
So, what to do with all this. In the end, I think it's best to take the time to build a high-quality PEC curve and try using it with your mount. Use it by itself (with PHD2 using Hysteresis guiding) to see if you have improved the RA tracking or not. If you have, then try using PPEC on top of that to be sure you don't have the problems described in scenario 3.
Hope this helps,
Bruce