I was watching the NY Times site pretty closely and never saw them give Saccone a 62% chance. The reason they took the predictor down was because Westmoreland County (a Republican county) was not providing precinct-level results. Therefore the Times was not able to use its modeling from past elections to make a prediction for this one based on incoming precinct results.
If you go back and review the “Meter” readings you will clearly see the Saccone percentages vary from 52-62% chance of winning. Then..like waving a magic wand...the percentages were erased and gone.