SbbPerson

Reactivity with two of three cells effectively rules out the "antibody to a low incidence antigen" argument.
Is the supplier/manufacturer of the Screening Cells the same as the supplier/manufacturer of the panel ? If not, then I suspect formulation differences of the two products may be the answer, specifically pH of testing environment. 

Because I've seen so many of them………….
But actually if your panel is not in the same buffer then pH or ionic strength could be the culprit rather than the temperature.
Important to stress that these cold anti-Ms (in that they dont react strictly at 37°C) have no clinical significance.
If it happens again, you can try the following:
1.  Repeat the panel, incubating for 15mins at RT (on a Coombs card).  The results will be stronger in this case.
AND 2.  Repeat the screen in the following way.  Warm the cells to 37°C  (best just to use a small aliquot - you dont want to 'cook' the whole bottle). And put the Coombs card  (foil still on) in the incubator for 15mins. Put the patient's plasma in the incubator. IN the incubator, pipette 50ul of cells into the appropriate wells followed by the patient's plasma.  Incubate for 15mins.  At the same time, start the centrifuge empty.  This warms the centrifuge up a bit.  After 15mins put the incubated card into the centrifuge which will have now stopped and centrifuge immediately.
This should negativise the screen results.  This is about the nearest you can get to doing a gel test at strictly 37°C

Thank you for replying.  Everything was done by gel card, we don't use the tube method.  Patient was typed A Pos. The antibody screen was 1+ positive on Screening cell 1 and 3, negative on Screening cell 2.  But when we did the Antibody Identification panel and Auto control , they were all negative. We repeated everything and got the same results.  We redrew the patient for a new specimen and got the same results. 
We sent out the specimen for an Antibody Identification at a reference lab. The result we got back was Anti-M. 
I am just wondering, why didn't our panels picked this up? Why only the screening cells were positive? What was the reasoning behind this? The panel and all our reagents have passed QC and are not expired. There was no signs of contamination in our specimen or reagents. Thank you. 

In addition to what has been nicely explained by Malcolm, it could be as well an example of Sd(a++) cell (commonly named "super Sid") reacting with a weak anti-Sda. The Sda antigen is not a LFA (expressed on more than 90% of cells) though some cells "overexpresse" it.
Anti-Sda usually gives weak/DP reactions and can be neutralized using urine (contains soluble Sda substances). 
Other weak antibodies may behave the same way, e.g. anti-P1 reacting against "strong P1" cells only.
However, that does not change at all what Malcolm said "I wouldn't expend too much time or energy trying to sort out the exact specificity.  In all cases of such an antibody, as long as you cross-match by the same method as you used in detecting the presence of the antibody in the first place, it would be quite safe to give cross-match compatible blood."

I'd parrot the same sentiments listed above. As long as you have ruled-out all other clinically significant antibodies, call the identification a low frequency and call it a day. You'll have to perform any future crossmatches with the same methodology. The chances of it being some crazy rare antibody is, of course, low.
There was also another thread listed here with a similar question.

This is a long shot, but if your patient is black and transfused often, it could be ant-Jsa or anti-V/VS.  We see that not infrequently with our sickle cell patients who are on a chronic transfusion protocol. 

It is a long shot.  I don't think I've ever seen a Jsa+ screening cell.
Plus - it would be nice to know how strong the reaction(s) was and which cell(s) was reacting.

It does occur, but what we see more often is a negative screen and about a quarter to a third of the units incompatable by Coombs crossmatch.  These patients get transfused about once a month, so we still do AHG crossmatches on them.