I agree with Michael_M on the customer possibly not understanding what they are looking for on the print.
The true position tolerance called out could possibly be done using a dial indicator but not with very good repeatability. First of all they have switched the datum as Claes has already pointed out. Second, the readings from the dial indicator using the setup the customer describes are telling you more about wall thickness which is more like a profile tolerance than true position tolerance. Profile tolerance can give you true position, but it is more restrictive. For example, in manufacturing your part could be made up of two ovals with their major axis at 90? to each other and still be within the true position tolerance. However they would fail at this measurement method that is giving you profile tolerance.
Using the indicator you could try to account for this by recording the indicator reading at set points around the circumference of the part and calculate true position with the addition of an outside diameter measurement of the part at the respective points measured using the indicator. Or you could just have this done automatically on a CMM.
Measuring runout requires keeping your part stable with respect to an axis of rotation. Rotating a cylinder on v-blocks does not keep the axis of rotation in the same position unless the cylinder is perfectly round. So you really are not measuring runout using the recommended setup either.
It is probably time to sit down with the customer and determine what it is they are really looking for from your product. If they want to ensure control of the size of the outside diameter with respect to the inside diameter they can continue to use the true position tolerance, but they need to find a more appropriate measurement method. Another alternative would be to change the tolerance to a concentricity tolerance, but again the measurement method would need to be changed to suit the tolerance. If they change to a profile tolerance the recommended measurement method could be potentially be kept but there is some cosine error that would need to be accounted for.
Isn't GD&T fun?