The old method overestimated your percentages by about 15%. The new estimate typically underestimates by 1% with a narrower range of error. Much better!
Sounds like you may have an issue with testing your true 1RM. The “moved like” estimate is based on your velocity. So the bar is not as close to failure as you think. This might be an indication to work with overloading variations more frequently before your next 1RM tests.
Yes, that’s where the trend points. The trick is there is usually some breakdown as you get closer. Wherever it breaks is where your tested 1RM is. 1RM testing is a skill too besides just a show of strength. Partial ROM movements are especially helpful as are reverse bands.