Thank you so much for the update! I have been monitoring the situation and really appreciate that you took the time to let us know. And thank you for the service you provide!

Thank you so much for the update, Michael!

I see that the times across the website and downloads are now consistent with each other, and seem to match the "new" set of parameters, where the times slowly start to get later and later, up to 3 minutes in August and September, and then the gap  between the new candle-lighting times and the original candle-lighting times narrows back down. 

When I compare the Hebcal times to the OU Zmanim, the new ones are the same or 1 minute different in January, but 2-3 minutes different in September. In comparison, the Hebcal times that were generated a year ago were consistently the same or only 1 minute different than the OU Zmanim.

What is the reason for the new seasonal difference?

Thank you so much for this resource, and for responding and updating us as to the status of the upgrades taking place! 
Meanwhile, which times are accurate? Are the times on the website correct? Or are the times in the PDF correct? Or even the subscribed calendars, as shown below?

While differences of 1 or 2 minutes may be expected, it's unclear which set of times is being generated by the updated/most accurate algorithm - or if none of them is currently reliable. If you can let us know which set of times to use right now, it would be extremely helpful. 

Thank you so much!