Motor efficiency after rewinding, Reduction & Improvement

As you know, motor rewinding will reduce motor's efficiency from %10 up to %15. Despite of that, efficiency may improve up to 5% reduction after some time working.
Now, my question is that,
1- why this reduction happen?
2- why this improvement happen?

I am not sure of the 5 % improvement unless this is a relative amount as in, say, 5 % improvement on 85 % is 1.05 x 85 % = 89.25 % which is still too big an improvement to be attributed to a motor 'settling in'.
The phenomenon I know which causes an improvement (reduction) of iron losses is annealing stress-relief of the punching-induced stresses to the structure of the electrical steel. The burnout procedure for the rewind will anneal the steel and hence reduce these additional losses. BUT I would not expect more than a percent or so in relative efficiency.
The other possibility is that the rewinder had managed (typically on older motors) to fit considerably more copper in the slot, to improve the efficiency. Even here a 5 % increase is beyond what I would believe is practical.
In all these cases - I would expect between 0.5 and 1.5 % improvement in efficiency - and these are unusual cases - but I believe the EASA study confirms this.
Most importantly - it concludes that with good procedures and tests and processes - even high efficiency motors can be - and are - rewound maintaining the original efficiencies to within standard tolerances.