Decompilation incorrectly uses pointer instead of the value it points to

Running RetDec built from https://github.com/avast-tl/retdec/commit/d57764abb7e45de61a22e45df298161b280316e1. I'm looking at a function that starts like this in the `.dsm` file:

```
0x10049390: 55 push ebp
0x10049391: 8b ec mov ebp, esp
0x10049393: 83 ec 14 sub esp, 0x14
0x10049396: 53 push ebx
0x10049397: 56 push esi
0x10049398: 8b 71 04 mov esi, dword ptr [ecx + 4]
0x1004939b: 8b da mov ebx, edx
0x1004939d: 57 push edi
0x1004939e: 8b 39 mov edi, dword ptr [ecx]
0x100493a0: 3b fe cmp edi, esi
0x100493a2: 0f 84 81 00 00 00 je 0x10049429
```

In the `.c` file I get the following:

```
int32_t function_10049390(int32_t a1, int32_t a2) {
// 0x10049390
int32_t v1; // bp-24
int32_t v2 = &v1; // 0x10049393
int32_t v3 = g4; // 0x10049396
int32_t v4 = g8; // bp-32
int32_t str = g5; // 0x10049398
int32_t v5 = *(int32_t *)(str + 4); // 0x10049398
g8 = v5;
int32_t v6 = g7; // 0x1004939b
int32_t v7 = g6; // bp-36
g6 = str;
if (v5 == str) {
```

So `ecx` is turned into `g5`, `edi` into `str` and `esi` into `v5` - so far it all makes sense. However, unlike indicated in the C program, `edi` and `ecx` are *not* the same. Rather, `ecx` points to a structure with two pointers, string start and string end. This appears to be some optimization gone wrong. I would rather expect the following here:

```
int32_t v5 = *(int32_t *)(g5 + 4); // 0x10049398
int32_t str = *(int32_t *)g5; // 0x1004939e
```

Reproduce the report with RetDec at commit d57764abb7e45de61a22e45df298161b280316e1, using the shown .dsm function and comparing its generated .c output. Trace how the ecx-based structure fields become edi and esi, then verify that the generated C dereferences both structure members and matches the assembly.

Written by the indexing model from the issue text.