Note that the OP said the images solve in ASTAP if loaded directly. When NINA passes an image to ASTAP, at least as far as I am aware, it reads the image and rewrites it, it is not just passing it on as a file pointer. The error as described sounds like it is failing when it tries to read the image, so changes in bin and scale will not help.
I think as the OP suggested capturing the image to solve in NINA will resolve the issue.
As to bin:
Note also there are two rescalings that can occur, one is by setting downsample on the platesolve -- this affects only new images NINA takes for plate solving. My recommendation is to leave that the same as you use for most of your light images, so if your normal images taken are bin 1, keep your plate solve at bin 1. The reason for this is that one of the most useful features of NINA's advanced sequencer is the center-after-drift, which helps recover from cloud (or other) induced offsets. It by definition is going to solve the image at your normal bin, so might as well have everything in the plate solve pipeline already binned that way. It is not necessary to leave it the same, if you follow the next recommendation, but it does mean your plate solves are from consistent sizes and so any issues will show up (or be avoided) in both lights and plate solve specific captures.
The other setting is the downsample. You can force ASTAP to downsample the resulting image before it solves by setting this (platesolve options under ASTAP specifically, the one above is generally platesolve not astap specific). This occurs AFTER any downsampling you put in the plate solve bin. The current recommendation for that is zero, which means "let ASTAP decide".
But generally speaking if NINA is not able to read the image, all this is moot -- if the original problem stems from NINA finding the image unreadable, address that first either by capturing the image yourself, or find out why and fix the source