Could you please attach the result of xkbcomp :0 -xkb out.xkb ?

On my system, I see funny configuration:

    modifier_map Control { <LCTL> };
    modifier_map Shift { <LFSH> };
    modifier_map Shift { <RTSH> };
    modifier_map Mod1 { <LALT> };
    modifier_map Mod4 { <LALT> };
    modifier_map Lock { <CAPS> };
    modifier_map Mod2 { <NMLK> };
    modifier_map Mod5 { <LVL3> };
    modifier_map Mod1 { <RALT> };
    modifier_map Mod1 { <LWIN> };
    modifier_map Mod5 { <MDSW> };
    modifier_map Mod1 { <META> };
    modifier_map Mod4 { <SUPR> };
    modifier_map Mod4 { <HYPR> };

So Mod4 is assigned to LALT, SUPR and HYPR
Mod1 is assigned to LALT, RALT, LWIN and META.

May be, there is an issue with xkbcomp here?

Created attachment 22139 [details]
xkbcomp output

...
    modifier_map Control { <LCTL> };
    modifier_map Shift { <LFSH> };
    modifier_map Shift { <RTSH> };
    modifier_map Mod1 { <LALT> };
    modifier_map Mod4 { <LALT> };
    modifier_map Lock { <CAPS> };
    modifier_map Mod2 { <NMLK> };
    modifier_map Mod5 { <LVL3> };
    modifier_map Control { <RCTL> };
    modifier_map Mod1 { <LWIN> };
    modifier_map Mod5 { <MDSW> };
    modifier_map Mod1 { <META> };
    modifier_map Mod4 { <SUPR> };
    modifier_map Mod4 { <HYPR> };
...

And xmodmap gives:
$ xmodmap 
xmodmap:  up to 3 keys per modifier, (keycodes in parentheses):

shift       Shift_L (0x32),  Shift_R (0x3e)
lock        ISO_Next_Group (0x42)
control     Control_L (0x25),  Control_R (0x69)
mod1        Super_L (0x40),  Alt_L (0x85),  Meta_L (0xcd)
mod2        Num_Lock (0x4d)
mod3      
mod4        Super_L (0x40),  Super_L (0xce),  Hyper_L (0xcf)
mod5        ISO_Level3_Shift (0x5c),  Mode_switch (0xcb)

That's really amazing.
$ grep Mod4 out_swap.xkb 
    modifier_map Mod4 { <LALT> };
    modifier_map Mod4 { <SUPR> };
    modifier_map Mod4 { <HYPR> };

And at the same time:
$ xmodmap | grep mod4
mod4        Super_L (0x40),  Super_L (0xce),  Hyper_L (0xcf)

My guess would be that it is xkbcomp's fault.

Created attachment 30517 [details] [review]
Fix option altwin:spap_lalt_lwin

setting explicit ONE_LEVEL for LALT seems to fix the bug

ok, if you're happy with that.

committed.