I can pretty much echo the experiences of @Dave Pickles.
I am using accurate recording with a UK Freesat setup. I have EIT, Freesat and Freesat (EIT) grabbers enabled. My recording profile has an extra warm up time of 300s and zero pre/post recording padding (I found that anything else for padding gave some strange results). I record everything using accurate recording and it's pretty solid on the main five UK channels (Channel 4 is the worst but this is pretty much always down to the broadcaster messing up the start/stop signals rather than TVHeadend). I also have a Freesat branded PVR (Humax Foxsat) and the results in terms of start/end recording times are identical between TVHeadend and the branded box.
The lesser channels available on Freesat are very variable in terms of whether or not the broadcaster provides the required identifiers in the EPG (Series and Episode CRIDs) and start/stop signals (EIT p/f packets in the transport stream) and, even if they do, whether or not these are transmitted at the right time. I spent a while a few years ago verifying this both in the TVHeadend logs and by comparison with identical recordings set on my branded receiver.
I occasionally have the odd recording that TVHeadend misses and I sometimes get additional episodes recorded from another series of the same programme. I believe (although I have never proven) that this is because a TVHeadend Autorec does "other things" to identify what to record (like matching on title) as well as matching purely on the Series CRID and Episode CRID in the EPG data. In my opinion, if using accurate recordings, the recording schedule (i.e. what to record) should be based solely on Series/Episode CRID and nothing else should be considered. Never had the time to delve into this though and prove whether or not my theory is correct.
I can also confirm that everything goes very wrong if you also enable a Freeview tuner and EPG (effectively the same as @Dave Pickles but the other way round). I also spent some time looking at this and came to the same conclusion in terms of Event IDs in the Freesat/Freeview EPG data. I had grand plans at one point to try to fix this and spent many hours tracing and delving into the code but, unfortunately, life got in the way and I've never had a chance to look any deeper.
In summary, I'd say it works with about 95+% success for me (but I'm only recording from the main five UK channels).