Context sensitive tokenizer misses certain kinds of time expressions

Expressions like:
4:39PM are not parsed correctly by the time finite state machine. Examples in the comments are:

• <li>3 PM</li>
• <li>3 P.M.</li>
• <li>3:05 PM</li>
• <li>3:05 P.M.</li>

But if there is no space between the minutes and PM the expression is not recognized.