Your initial investigation is moving in the right direction. The best way to trouble shoot this type of problem is to isolate the three parts, board, component and solder paste. Since you have reviewed the solder paste and found it to be acceptable and perhaps even tried another lot, I would focus on the boards and components.

With the problem isolated to a couple of component types you could almost eliminate the board and check the solderability of the components. One simple test is print the component stencil pattern of the suspect component on a ceramic substrate. Next place the component and reflow as usual. After reflow you can easily remove the component and inspect for solder wettability. If it's unacceptable you now have evidence to push back on your component supplier for help.

Dewetting on specific areas of the board can be from several sources. Dewetting typically is caused by a solderability issue related to a passive surface, this sometimes is under the gold or HALS or other coating. It can be from intermetallics or oxides. Time above liquidus also can cause this, excessive time above liquidus will cause the solder to pull back. Is this a heavy ground area that has a different temperature gradient? Is there a specific surface finish on the components that are dewetting?