Strange behaviour M61 Qn and cuttercompensation

Because i need to do manualtoolchange sometimes at other position than at defined toolchange position, i use M61 Qn instead of M6. Works fine, mostly. Only this situation example:
Having a flute mill dia=12mm in spindle, next tool is dia=2mm, and cutter compensation will be applied. When reload file and M61 Qn is used for toolchange, it occurs an error in connection with G41/G42. Cuttercompensation does incorrectly recognise current tool in spindle, instead of changed tool. Is there a way to make M6 toolchange position variable, setable from Gcode, or get rid off this behaviour of M61 Qn / G41 G42?

I use tooltable, and G43 is always applied. I made a short .ngc file to demonstrate, only need to load and reload. Tools T12 D=12 and T2 D=2 needed in tooltable.
Comment/uncomment in line 30/31. Open file or Ctrl-R produces the error, when M61 Q2 used.

After this problem occurred again yesterday, i remember to bump it back up the list.

Thank you for investigating about this problem. Somehow i understand the description, but i am very far from being able to find a solution myself. As i completed this years work, I'm away for a while anyway.

Now i circumvent this problem by Remap M6. The initial demand was, to be able to goto a freely selectable (manual)toolchange and probing position. I got this option, by removing toolchangeposition related entrys in the ini, adding variables for TCH and Z_OFFS, the standard toolchange and probing position. I have also the option in the .ngc file, to set this positions. If there is an entry in the .ngc, it overrides the .ini entry. At the same time, tool and z offset probing was improved. The first tool in the program sequence gets probed, then the toolsetter is placed on the workpiece, to get Z-offset, where a variable can be applied for minor corrections of the probing result. As we know, the toolpath doesn't match the preview anymore, after probing. So my probing routine has some funny testcounter, which allow to abort the program, ctrl-r, and when the program is started again, the 2 probing actions are skiped once, and the counter reset. A toolchange will reset this option also. This is only for the very beginning of a routine, with the first tool.
I did the testing in the simulation, also used the simulate_probe button, but still was tired of g38.n needing too much attention. A .ini entry 'SIM' with different values for the machine and simulation was made, to exclude certain programsequences in the simulation.