The question of pattern is an open one (and likely not a conclusive one i.e. infinitely many patterns are equivalent), but the use of patterned paste is a well trusted and documented one. Here's an application note for example:
Using screen-printing templates to apply thermal
grease in IGBT modules | Infineon
A toothed trowel could also be used, making a grooved surface (see: tile mortar / mastic trowel). The grooves need to be fairly shallow, of course; not like a hardware store trowel!
(Although come to think of it, maybe you could spread an even layer of paste to start with, then use such a trowel to scrape furrows out. Repeat crosswise if a square-dot pattern is desired. Would be a bit messy, but could be an option...hmm!)
The resulting furrows might trap air, but those voids should never be wider than the groove spacing.
There is still the question of permissible voids. This depends on size of the void, thermal conductivity of the aluminum PCB, and power level. Heat can flow sideways through the PCB to nearby grease spots, raising the temperature of that source (i.e. a source directly above such a void) by some nominal amount. It might well be that your power level is low enough that it's highly noncritical, and a bit smeared on here and there would do.
Although I would guess a PCB in free air would do well enough in that case, not needing a heatsink at all. Perhaps your power level is pretty aggressive, on the order of a kW say; in that case, you may want to take thermal measurements and verify that you're getting the performance you require.
Keep in mind also, the grease only smooshes down so far, at some distance from the screws. How much, depends on the stiffness of the board (so, thickness and material). Paste is only a modest conductor by itself, with its main value being to displace thin layers of air (which is a particularly poor conductor in thin gaps). A full calculation might involve device size and power, and the conductivity and thickness of the PCB, grease, and heatsink; mind that the sink itself need not have a uniform temperature, so is worth including in the calculation/simulation. (A full simulation is probably more effort than it's worth; just try a few paste strategies and see what measures best. Do apply power gradually, so as to avoid overheating things in case a setup performs poorly.)
There's also the option for gap-filling pads, which are made from a tacky rubber material with good thermal conductivity. Self-adhesive materials are also available, providing a near-permanent bond between pieces. Since a PCB of this size, and average thickness, will be modestly flexible, such an assembly could be rolled together without too much trouble. (Board flex can take up larger-scale errors in flatness, and the pad takes up smaller-scale errors or surface roughness.)