Semantic Kernel: The integration of chat and text completion services into the context does not utilize the models as expected

Question

I'm developing an Azure Function and I've integrated SK to it injecting the kernel as seen in many starter projects.

//Microsoft Semantic Kernel configuration build
var skBuild = Kernel.Builder
                    .WithLoggerFactory(loggerFactory)
                    .WithAzureTextEmbeddingGenerationService("text-embedding-ada-002", azureOpenAIOptions.Endpoint, azureOpenAIOptions.ApiKey)
                    .WithAzureTextCompletionService("gpt-35-turbo-instruct", azureOpenAIOptions.Endpoint, azureOpenAIOptions.ApiKey)
                    .WithAzureChatCompletionService("gpt-35-turbo", azureOpenAIOptions.Endpoint, azureOpenAIOptions.ApiKey)                        
                    .WithMemoryStorage(memoryStore)
                    .Build();

Even I specified different models for each searvice

.WithAzureTextCompletionService("gpt-35-turbo-instruct",... ,...)

and

.WithAzureChatCompletionService("gpt-35-turbo",... ,...)

SK is only using the model specified for chat completion service. What I'm expecting is to SK uses the correct model depending on implementation.

For example: gpt-35-turbo-instruct when implementing

var result = await kernel.RunAsync(context, skill["Joke"]);
    Console.WriteLine(result);

and gpt-35-turbo when implementing

var chat = chatCompletionService.CreateNewChat("You are an AI assistant that helps people find information.");
        chat.AddMessage(AuthorRole.User, "Hi, what information can you provide for me?");
    
        string response = await chatCompletionService.GenerateMessageAsync(chat, new ChatRequestSettings());
        Console.WriteLine(response);

I'm missing something?