Semantics of different abstract implict cast kinds are not the same

[Gama11]

https://haxe.org/manual/types-abstract-implicit-casts.html claims that to and @:to should be identical in behavior:

Class field casts have the same semantics, but are defined completely differently:

In practice this doesn't seem to be the case:

import haxe.ds.HashMap;

class Main {
	static function main() {
		var map1 = new HashMap<Hashable1, Bool>(); // works
		var map2 = new HashMap<Hashable2, Bool>(); // constraint check failure
	}
}

abstract Hashable1(HashableImpl) to {function hashCode():Int;} {}

abstract Hashable2(HashableImpl) {
	@:to function toHashCode():{function hashCode():Int;} {
		return this;
	}
}

class HashableImpl {
	public function new() {}

	public function hashCode():Int {
		return 0;
	}
}

@RealyUniqueName says this is by design, and thus should be documented accordingly. Would be helpful to have a list of situations where @:to and @:from are not considered.