This is an interesting article, but the cross-validation is perhaps suspect? It seems likely the 'percentage used' indicator is based on the same underlying mea

This is an interesting article, but the cross-validation is perhaps suspect? It seems likely the 'percentage used' indicator is based on the same underlying measurements as the 'estimated capacity' service indicator. So it's quite likely illusory validation of the 4% loss figure.