The invalid document can't be measured and IndexOutOfRangeException/NullReferenceException will be thrown.

I have tried to use the RichTextBox, but found that it does not work very well when it comes to measure the text. This results in clipped text and that the caret can be displayed in a different place then expected.
This can easily be seen with a font like Gabriola, which uses complex kerning metrics. Type a text with a lot of "ff" and "fi" and you will soon see that your caret gets out of sync and that text is sometimes clipped.
When I look at the code of RichTextBox it seems to me that it uses its own calculation of text widths and that the this is based on a very simple method using just the kerning information. There is also a hard coded list containing information which fonts that has kerning info and which fonts that are mono spaced. It seems strange to me to have this in a config file since there is a large number of fonts used by the users.
It also seem to look at the Unicode character family in order to determine if ligatures are used or not, which is also quite strange since ligatures can be used for western characters too. The font Gabriola is an example of this. Look at "fi" in a word like "fine".
It can potentially consume a lot of memory since I think that it builds a dictionary of the kerning of all pairs of characters for all fonts that it encounters.
I guess this way of doing things were made for performance, but unfortunately, good performance doesn't help much if it doesn't work.
I am not sure if I should report this as a bug, since it is more of a fatal design issue of the whole control.

please add shapes support for radrichtextbox. else support another textbox

Please look for the same topic in the forum for context. 
@Todor:
Am attaching the wmf file and png file that don't get rendered.

Possible workaround is to change the document of RadRichTextBox.

The fix will be available in our next LIB release (v. 2016.1.208).

Tab leader (dots) are calculated based on the available space between the end of the text in the entry and the page number. 

Workaround: to change rendering mode to TextBlockWithPropertyCaching:

this.radRichTextBox.TextRenderingMode = TextBlockRenderingMode.TextBlockWithPropertyCaching;

Workaround: Caret could be changed with custom implementation. More information could be found in this help article: http://docs.telerik.com/devtools/wpf/controls/radrichtextbox/ime-support#how-to-implement-your-own-ime-support

When a color property has default (auto) value, do not export a tag which points to omitted defintion in the color table. For example, if a span has transparent highlight (the default one), do not export the \highlight tag since this might lead to some incorrect rendering of the document from other readers - WordPad for example.

The issue is sporadic. Updating the UI resolves it.

Workaround: manually call
this.radRichTextBox.ActiveEditorPresenter.RecreateUI();
just after the acception.

Deleting a row in the attached table causes a crash during document layout.

It seems the height of the line is not properly calculated for specific fonts which leads to difference in the way the documents look. 

Document with one full page of text formatted with Arial Unicode MS looks shorter in RadRichTextBox.

When ParagraphProperties dialog is opened while several paragraphs are selected, the values in the dialog correspond to the values of that paragraph, where the cursor was placed before opening the dialog. Then if the dialog is closed with OK button, these values are applied to the other paragraphs as well.

In specific conditions, when selecting text through the document, NullReferenceException is thrown in RectPathBuilder class.

A "FontSize" setting cannot be applied to multiple paragraphs when the font size of the character at the beginning of selection is equal to the value which is going to be set.
Update: the issue is duplicate of http://feedback.telerik.com/Project/143/Feedback/Details/173836 .
The fix will be included in 2016 Q1.