Order of operations in mxAlgebra()'s

I had thought that there would be a default order of operations of mxAlgebra() operators. Indeed, consultation of the wiki http://openmx.psyc.virginia.edu/wiki/operator-precedence seems to confirm that binary + should be done after %*%
However, the following example suggests that this order is not followed. The only difference between the equations is that the two parts separated by the binary + have been offset by (). Since this is lower precedence than the other operations, it should have been done last (the binary - should have been done earlier inside the solve() function call). But different results emerge.

> mxEval((solve(ACE.I-ACE.at) %&% (ACE.ai %*% t(ACE.ai)) + (ACE.fl %&% (ACE.alp %*% t(ACE.alp)))), TraitStateACEFixFl1stModelFit)
[,1] [,2] [,3] [,4]
[1,] 0.4580807 0.4580807 0.4580807 0.4580807
[2,] 0.4580807 0.4805261 0.4504615 0.4625746
[3,] 0.4580807 0.4504615 0.4609001 0.4564178
[4,] 0.4580807 0.4625746 0.4564178 0.4593563
> mxEval(solve(ACE.I-ACE.at) %&% (ACE.ai %*% t(ACE.ai) + ACE.fl %&% (ACE.alp %*% t(ACE.alp))), TraitStateACEFixFl1stModelFit)
[,1] [,2] [,3] [,4]
[1,] 0.4580807 0.3635496 0.3346717 0.2606875
[2,] 0.3635496 0.3109715 0.2579885 0.2113850
[3,] 0.3346717 0.2579885 0.2473290 0.1887942
[4,] 0.2606875 0.2113850 0.1887942 0.1496293
>