Teasing up +4.5 dogs may have worked in the past, but its not the way to break down teasers. You are capping the future, not the past.
Look at it this way.
A Game lined -4.5 will have a moneyline of around -215/+195, which translates into 33.2 outright wins by the dog, and 66.8 outright losses.
Now in order to get the teaser % into the 72 -73 % range, that would mean that the dog would have to stay withing 10.5 points in about 39 of the 66.8 games that it will lose, about 58.3 % of the time.
Do a simple analysis on avg MOV by spread, break it down by home/road, AFC/NFC etc, look at how often teams cover +10.5 WHEN THEY LOSE, you will see that this subset is no good. Even when 2 and 3 point dogs lose, they dont stay within 10.5 points of the final score 58 % of the time. If 2 and 3 point dogs can't hit that #, then surely 4.5 dogs won't.
For fun I drilled down on my database which covers 1978-present.
Dogs of 4 and 4.5, when they lose, stay within 10.5 points of the final score only about 53 % of the time.