That's not true. The comment column is for component values.

@BenRoe that table is from JLC's web site and is on the page I linked above. It is not at all clear from the commentary on that page how the various fields in the BOM are used in the fabrication process. My experience was that putting the MPN in the comment column gave much better results for part selection.

Note that the patch doesn't force new behavior. For my use I could achieve 100% correct part matching with JLC's system using the option compared to 80% failure with the default setting.

I've been having similar issues, as we move from the KiCad JLCPCB Tools plugin to Fabrication Toolkit. This section of our BOM highlights the issue:

Designator,Footprint,Quantity,Value,LCSC Part #
J3,HC-5566-2x4A-05,1,LOOM_A_OUT,C20539149
J4,HC-5566-2x4A-05,1,LOOM_A_RETURN,C20539149
J5,HC-5566-2x4A-05,1,LOOM_B_OUT,C20539149
J6,HC-5566-2x4A-05,1,LOOM_B_RETURN,C20539149

Even though the part number is specified, JLCPCB's ordering system seems to ignore it

Conversation with JLCPCB's support got me to this:

We have carefully reviewed the information. The behavior you encountered occurs because our system currently prioritizes matching components based on the "Comment" field in the BOM. In your file, the descriptive column is named "Value", which can sometimes cause the system to not immediately recognize and lock onto the "JLCPCB Part #" you specified as the primary choice, leading to the manual matching prompt. Additionally, when the same part appears on multiple lines in the BOM, the system flags it with a "(!)" alert as a standard check to confirm if the separate listings are intentional.

Manually tweaking the generated BOM file to change the column header from LCSC Part # to Comment does seem to fix it (counterintuitively). There's a solid argument to make that it's their systems not behaving as expected rather than needing a change here I guess. I'll fix our files in post-processing for now, but this change would get a +1 from me