Sorry for the delay. I had a vacation in Europe and couldn't access the forum.
Thank you for 1-wire microchip note. I haven't seen it before.
I guess the timing issue happens when reading from the sensor. My code pulls down, waits 3 us, then releases and samples in 10 us (so, I use 3/10, which is close to microchip 6/9). So, there's 10 us between PIC releasing the line and sampling. If it's 12 us to pull up against 1500 pF, then it is more than my 10 us, meaning it may not be pulled up when I sample. It works some of the time, but not very consistent. When I use 4.7K on the sensor end and 2.2K on the PIC end, I get 1.5K pullup, which works much better.
Looking as micriochip recommends it, I probably be better with 3/12 instead of 3/10, which will give it 2 us more to pull up.