I say live with it.
The roughly 18 degree temperature difference will only cause the journal to appear to be .0001 smaller than it actually would be if measured at the correct temperature. Actually, if the measuring equipment and its calibrating gage blocks were at a similar temperature they would also be that much smaller so the reading you got would be still be accurate.
As I mentioned before, the .0003 tolerance (1.0000-.9997) is for a brand new journal size. It is anticipated by the designer that there will be some wear during the course of the journal's life and these hardened steel journals can not be "built up" to enlarge them.
For that reason a limit lower than the original new size is allowed before scrapping the crankshaft would be required.
The service limit of .9994 I suggested in my previous post will effect the radial clearance in the bearing assembly by allowing it to be .00015 looser than a brand new journals lower limit. The oil film between the rollers and the journal will eliminate any perceptible play in the bearing even with this added clearance.
I'm assuming the journal that measured .9991 diameter is on the lightly loaded side of the crankshaft? If so, it should still work for many miles as long as the engine is not run hard.
As a side note, if this wasn't a surface with rolling element bearings running against it, it could be repaired by "thin dense chrome" or "electroless nickel" plating. These repairs could be used on a shaft that was going to have a full bearings inner ring pressed onto it, for instance.
These methods of repair cannot be used with your engines design because the direct contact between the rollers and the journal would rapidly chip off any plating that was applied to the crankshaft.
As I said, live with what you've got. Run the engine easy and it should last for many years.